Knowl. Org. 46(2019)No.8
KO KNOWLEDGE ORGANIZATION
Official Journal of the International Society for Knowledge Organization ISSN 0943 – 7444 International Journal devoted to Concept Theory, Classification, Indexing and Knowledge Representation
Contents
Vanda Broughton. Joachim Schöpfel, Dominic Farace, Hélène Prost and Obituary: Antonella Zane. Emeritus Professor Ia McIlwaine: Data Papers as a New Form of Knowledge An Appreciation ...... 573 Organization in the Field of Research Data ...... 622
Pablo Gomes and Maria Guiomar da Cunha Frota. Special Issue: Knowledge Organization from a Social Perspective: Best papers from NASKO, ISKO-UK, Thesauri and the Commitment to Cultural Diversity ...... 639 ISKO-France, ISKO-Brazil 2019 Mario Barité. Articles Towards a General Conception of Warrants: First Notes ...... 647 Heather Moulaison Sandy and Andrew Dillon. Mapping the KO Community ...... 578 Research Trajectories in Knowledge Organization
Elliott Hauser and Joseph T. Tennis. Barbara H. Kwaśnik. Episemantics: Aboutness as Aroundness ...... 590 Changing Perspectives on Classification as a Knowledge-Representation Process ...... 656 Vanda Broughton. The Respective Roles of Intellectual Creativity Sixth Annual “Best Article in KO Award” and Automation in Representing Diversity: for Volume 45 (2018) ...... 668 Human and Machine Generated Bias ...... 596 Books Recently Published ...... 669 Shu-Jiun Chen. Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China ...... 607
Viviane Clavier. Knowledge Organization, Data and Algorithms: The New Era of Visual Representations ...... 615
Knowl. Org. 46(2019)No.8
KNOWLEDGE ORGANIZATION KO
Official Journal of the International Society for Knowledge Organization ISSN 0943 – 7444 International Journal devoted to Concept Theory, Classification, Indexing and Knowledge Representation
KNOWLEDGE ORGANIZATION José Augusto Chaves GUIMARÃES, Departamento de Ciência da Informacão, Universidade Estadual Paulista–UNESP, Av. Hygino Muzzi This journal is the organ of the INTERNATIONAL SOCIETY FOR Filho 737, 17525-900 Marília SP Brazil. E-mail: [email protected] KNOWLEDGE ORGANIZATION (General Secretariat: Amos DA- VID, Université de Lorraine, 3 place Godefroy de Bouillon, BP 3397, Michael KLEINEBERG, Humboldt-Universität zu Berlin, Unter den 54015 Nancy Cedex, France. E-mail: [email protected]. Linden 6, D-10099 Berlin. E-mail: [email protected]
Editors Kathryn LA BARRE, School of Information Sciences, University of Illi- nois at Urbana-Champaign, 501 E. Daniel Street, MC-493, Champaign, IL Richard P. SMIRAGLIA (Editor-in-Chief), Institute for Knowledge Or- 61820-6211 USA. E-mail: [email protected] ganization and Structure, Shorewood WI 53211 USA. E-mail: [email protected] Devika P. MADALLI, Documentation Research and Training Centre (DRTC) Indian Statistical Institute (ISI), Bangalore 560 059, India. Joshua HENRY, Institute for Knowledge Organization and Structure, E-mail: [email protected] Shorewood WI 53211 USA. Daniel MARTÍNEZ-ÁVILA, Departamento de Ciência da Informação, Peter TURNER, Institute for Knowledge Organization and Culture, Universidade Estadual Paulista–UNESP, Av. Hygino Muzzi Filho 737, Shorewood WI 53211 USA. 17525-900 Marília SP Brazil. E-mail: [email protected]
J. Bradford YOUNG (Bibliographic Consultant), Institute for Knowledge Widad MUSTAFA el HADI, Université Charles de Gaulle Lille 3, URF Organization and Structure, Shorewood WI 53211, USA. IDIST, Domaine du Pont de Bois, Villeneuve d’Ascq 59653, France. E-mail: [email protected] Editor Emerita H. Peter OHLY, Prinzenstr. 179, D-53175 Bonn, Germany. Hope A. OLSON, School of Information Studies, University of Wiscon- E-mail: [email protected] sin-Milwaukee, Milwaukee, Northwest Quad Building B, 2025 E New- port St., Milwaukee, WI 53211 USA. E-mail: [email protected] M. Cristina PATTUELLI, School of Information, Pratt Institute, 144 W. 14th Street, New York, New York 10011, USA. Series Editors E-mail: [email protected]
Birger HJØRLAND (Reviews of Concepts in Knowledge Organization), K. S. RAGHAVAN, Member-Secretary, Sarada Ranganathan Endowment Department of Information Studies, University of Copenhagen. E-Mail: for Library Science, PES Institute of Technology, 100 Feet Ring Road, [email protected] BSK 3rd Stage, Bangalore 560085, India. E-mail: [email protected].
María J. LÓPEZ-HUERTAS (Research Trajectories in Knowledge Heather Moulaison SANDY, The iSchool at the University of Missouri, Organization), Universidad de Granada, Facultad de Biblioteconomía y 303 Townsend Hall, Columbia, MO 65211, USA. Documentación, Campus Universitario de Cartuja, Biblioteca del Colegio E-mail: [email protected] Máximo de Cartuja, 18071 Granada, Spain. E-mail: [email protected] M. P. SATIJA, Guru Nanak Dev University, School of Library and Infor- Editorial Board mation Science, Amritsar-143 005, India.
E-mail: [email protected] Thomas DOUSA, The University of Chicago Libraries, 1100 E 57th St, Chicago, IL 60637 USA. E-mail: [email protected] Aida SLAVIC, UDC Consortium, PO Box 90407, 2509 LK The Hague, The Netherlands. E-mail: [email protected] Melodie J. FOX, Institute for Knowledge Organization and Structure, Shorewood WI 53211 USA. E-mail: [email protected]. Renato R. SOUZA, Applied Mathematics School, Getulio Vargas
Foundation, Praia de Botafogo, 190, 3o andar, Rio de Janeiro, RJ, 22250- Jonathan FURNER, Graduate School of Education & Information Stud- 900, Brazil. E-mail: [email protected] ies, University of California, Los Angeles, 300 Young Dr. N, Mailbox 951520, Los Angeles, CA 90095-1520, USA. Rick SZOSTAK, University of Alberta, Department of Economics, 4 E-mail: [email protected] Edmonton, Alberta, Canada, T6G 2H4. E-mail: [email protected]
Claudio GNOLI, University of Pavia, Science and Technology Library, Joseph T. TENNIS, The Information School of the University of Wash- via Ferrata 1, I-27100 Pavia, Italy. E-mail: [email protected] ington, Box 352840, Mary Gates Hall Ste 370, Seattle WA 98195-2840 USA. E-mail: [email protected] Ann M. GRAF, School of Library and Information Science, Simmons University, 300 The Fenway, Boston, MA 02115 USA. Yejun Wu, School of Library and Information Science, Louisiana State E-mail: [email protected] University, 267 Coates Hall, Baton Rouge, LA 70803 USA. E-mail: [email protected] Jane GREENBERG, College of Computing & Informatics, Drexel University, 3141 Chestnut Street, Philadelphia, PA 19104 USA, E-mail: Maja ŽUMER, Faculty of Arts, University of Ljubljana, Askerceva 2, [email protected] Ljubljana 1000 Slovenia. E-mail: [email protected]
Knowl. Org. 46(2019)No.8 573 Obituary. Emeritus Professor Ia McIlwaine: An Appreciation
Obituary. Emeritus Professor Ia McIlwaine: An Appreciation
DOI:10.5771/0943-7444-2019-8-573
Professor Ia McIlwaine was born Ia Cecilia Thorold on 20 library schools, she kept it solidly and centrally on the cur- April 1935, the only daughter of Michael Thorold, a cler- riculum at UCL, and sustained it as a distinctive feature of gyman in the Church of England, and Dorothy Henfrey; the UCL department with a programme of international she had two younger brothers. Her unusual Christian events and a lively group of research students; some of her name, which often caused confusion for correspondents, doctoral students now occupy leading roles in the world of is Cornish in origin, the nominative form of what we may library and information science in general, and classifica- be more familiar with in the place name St. Ives. tion in particular. As a student myself in the early 1970s I She was a pupil at Bath High School, and, after leaving found the classification element of the course the most ap- school, went up to Bedford College, London to read Clas- pealing and intellectually engaging, and it led me into the sics, a discipline in which she maintained a lifelong interest. most rewarding career which I would not have enjoyed In 1957-58 she studied for the Graduate Diploma in Li- without her original enthusiasm and expertise. brarianship at University College London, the beginning Although her academic work was primarily focused on of a long and distinguished association with that institu- classification, where most of her publications are to be tion. After a five year period as Assistant Librarian with found, she had a broader interest in subject work and bib- Westminster City Libraries, during which time she was liography generally, and in bibliographic control. She co- awarded Fellowship of the Library Association, she was authored the book Introduction to Subject Study, and edited a appointed to the post of Lecturer in the School of Library number of conference and collected papers including & Archive Studies at UCL. A major part of her role there Standards for the International Exchange of Bibliographic Infor- was to teach classification and indexing, the subject field mation, Subject Retrieval in a Networked Environment, and which would become her primary research area. Knowledge Organization and the Global Information Society. Her She progressed steadily through the academic ranks, be- reputation as an editor led to the role of series editor for ing promoted to Senior Lecturer in 1985 and Reader in Saur’s (later de Gruyter) substantial Introduction to Information Classification and Indexing 1995, at which time she also Sources, which documented the bibliography of a wide num- took on the Directorship of the School, now the School of ber of disciplines and formats. On the classification front, Library, Archive & Information Studies. In 1997 she was she was an active member of the International Society for honoured with a Chair of Library & Information Studies. Knowledge Organization, being President from 2001-2005, She continued the strong tradition of classification and in- as well as a longstanding member of the Scientific Advisory dexing work in the School, which had begun with Charles Committee, and the Editorial Board of its journal Knowledge Berwick Sayers, whose pupil Ranganathan was in the 1920s. Organization. She was also, in its latter years, Secretary of the At a time of waning interest in classification in most UK UK Classification Research Group. In addition, she was a 574 Knowl. Org. 46(2019)No.8 Obituary. Emeritus Professor Ia McIlwaine: An Appreciation member of the British Standards Institution Committee on Committee from 2001-2003. She was also a member and Indexing, an observer on the Committee of the Bliss Clas- office holder in the Section for Classification and Indexing sification Association, and served on the British Committee and the Division of Knowledge Management, and in 2005 for the Dewey Decimal Classification. was awarded the IFLA Medal, “for distinguished services In what was probably her major contribution to the to IFLA.” field, she took over in 1993 as Editor-in-Chief of the Uni- Much of this activity continued into her early years of versal Decimal Classification, leading a major programme of retirement, but her time was increasingly spent at the family revision of this large international system. A major part of cottage in Norfolk, and later at a larger house where she the revision work under her Editorship was the introduc- could indulge her love of gardening. After a period of ill tion of a more rigorous analytical and faceted approach to health she died on 24 August 2019 from complications fol- the classification as a whole, which resulted in a number of lowing pneumonia. In 1966 she had married her fellow lec- radically revised classes, as well as a systematic pruning and turer, John McIlwaine, who was her colleague and her com- rationalisation of the scheme overall. To this end, she en- panion for 53 years. He survives her, together with their gaged the help of a number of colleagues and contributors, daughter Anne who followed her parents into the world of with whose assistance several substantial revisions of main libraries. classes were achieved. She undertook much of this pains- Ia was tireless in her work for the School of Librarian- taking and time-consuming work herself, producing new ship, for University College, and for the wider world of li- schedules for photography and large parts of the auxiliary brarianship, never shirking a difficult situation, and always table for “place,” as well as a major overhaul of the medi- going the extra mile. It was often her personal involvement cine class, carried out in association with Nancy William- and attention to detail that ensured the success of any num- son. ber of activities, both at home and abroad, including the She gave much thought to the problems of maintenance hosting of several international conferences on classifica- and revision of large classification systems, particularly tion and information retrieval. A sometimes daunting man- against the background of transition to an online environ- ner hid a well of personal kindness, and a considerable ment. Within that context, another innovative feature of sense of responsibility. She was a formidable administrator, her time as editor of the UDC was her fostering of a more but also a great enabler of younger colleagues and associ- cooperative and collaborative view of classification revi- ates who widely acknowledge the role she played in encour- sion. She worked closely with the editors of both the Dewey aging and advancing their roles in the profession. It was a Decimal Classification, and the Bliss Bibliographic Classification, not uncommon occurrence for some overseas visitor to both to share the effort of maintaining these large systems, stroll into the School Office at UCL, uninvited and unan- and also to promote consistency in the way they repre- nounced (and usually on a Friday afternoon), confident that sented subject content. she would welcome them, which she invariably did. A vast Alongside her work in classification and knowledge or- network of connexions across the libraries of the world ganization she retained her interest in the classics. Her PhD bears testimony to the friendship and respect she was ac- work, published as Herculaneum: A Guide to Printed Sources, corded by her fellow practitioners, as well as her academic was also carried out at UCL while she was a member of contacts in all countries. Her part in the international scene staff, and was awarded the prestigious Dunn & Wilson can hardly be exaggerated, and she will be much missed by prize. This interest was picked up again in retirement, when her friends and colleagues across the globe. she produced a supplement to her thesis for a new publica- tion from Bibliopolis for Centro Internazionale per lo stu- Vanda Broughton dio dei papiri ercolanesi. For this scholarly work she was Emeritus Professor of Library & Information Studies, elected to a Fellowship of the Society of Antiquaries, an University College London honour which particularly pleased her.
McIlwaine, I. C., John H. McIlwaine and Peter G. New, McIlwaine, I. C. 2000. “Interdisciplinarity: A New Re- eds. 1983. Bibliography and Reading: A Festschrift in Honour trieval Problem?” In Dynamism and Stability in Knowledge of Ronald Staveley. Metuchen, NJ: Scarecrow. Organization: Proceedings of the Sixth International ISKO McIlwaine, I. C., ed. 1991. Standards for the International Ex- Conference, Toronto, Canada, July 10-13, 2000, ed. Claire change of Bibliographic Information: Papers Presented at a Beghtol, Lynne Howarth, and Nancy J. Williamson. Ad- Course Held at the School of Library, Archive and Information vances in Knowledge Organization 7. Würzburg: Er- Studies, University College London, 3-18 August 1990. Lon- gon, 261-267. don: Library Association. McIlwaine, I. C., ed. 2003. Subject Retrieval in a Networked En- vironment: Proceedings of the IFLA Satellite Meeting Held in Classification and information retrieval Dublin, OH, 14-16 August 2001 and Sponsored by the IFLA Classification and Indexing Section, the IFLA Information Tech- McIlwaine, I. C. 1983. “Subject Analysis and Subject Study.” nology Section and OCLC. Munich: Saur. In Bibliography and Reading: A Festschrift in Honour of Ronald McIlwaine, I. C. 2003. “Trends in Knowledge Organization Staveley, ed. I. C. McIlwaine, John H. McIlwaine and Peter Research.” Knowledge Organization 30: 75-86. G. New. Metuchen, NJ: Scarecrow, 118-30. McIlwaine, I. C., ed. 2004. Knowledge Organization and the McIlwaine, I. C. 1993. “Subject Control: the British View- Global Information Society: Proceedings of the Eighth Interna- point.” In Subject Indexing: Principles and Practices in the tional ISKO Conference, London, England, July 13-16, 2004. 90's: Proceedings of the IFLA Satellite Meeting Held in Lis- Advances in Knowledge Organization 9. Würzburg: Er- bon, Portugal, 17-18 August 1993, ed. R.P. Holley et al. gon. Berlin: De Gruyter, 166-80. McIlwaine, I. C. 2010. “Universal Bibliographic Control and McIlwaine, I. C. 1995. “Preparing Traditional Classifica- the Quest for a Universally Acceptable Subject Arrange- tions for the Future: Universal Decimal Classification.” ment.” Cataloging & Classification Quarterly 48, no. 1: 36-47. Cataloging & Classification Quarterly 21, no. 2: 49-58. McIlwaine, I. C. and Vanda Broughton. 2000. “The Classifi- Mcllwaine, I.C. 1996. “New Wine in Old Bottles: Problems cation Research Group: Then and Now.” Knowledge Or- of Maintaining Classification Schemes.” In Knowledge ganisation 27: 195-99. Organization and Change: Proceedings of the 4th International McIlwaine, I. C. and N. J. Williamson. 1999. “International ISKO Conference, Washington. 15-18 July 1996, ed. by R. Trends in Subject Analysis Research: An Edited and Up- Green. Advances in Knowledge Organization 5. Frank- dated Version of a Presentation Made at the 1998 Annual furt/Main: Indeks Verlag, 122-36. Meeting of the American Society for Information Sci- McIlwaine, I. C. 1997. “Classification Schemes: Consulta- ence (ASIS).” Knowledge Organization 26: 23-29. tion with Users and Cooperation Between Editors.” In Plassard, M., F. Bourdon, M. Witt and I. C. McIlwaine. Cataloging and Classification: Trends, Transformations, Teach- 1998. “IFLA Core Programme for Universal Biblio- ing, and Training, ed. J. R. Shearer and A. R. Thomas. graphic Control and International MARC (UBCIM) New York: Haworth, 81-95. [Also published in Catalog- and Division of Bibliographic Control: Reports on Ac- ing & Classification Quarterly 24, nos. 1/2: 87-90.] tivities 1997-1998.” International Cataloguing and Biblio- McIlwaine, I. C. 1997. “Classifications and Linear Orders: graphic Control 27, no. 4: 63-67. Problems of Organizing Zoology.” In Knowledge Organi- zation for Information Retrieval: Proceedings of the 6th Interna- Universal Decimal Classification tional Study Conference on Classification Research, London, 16- 18 June, 1997, ed. J. H. Bowman. The Hague: FID, 134- McIlwaine, I. C. 1991. “Present Role and Future Policy for 38. UDC as a Standard for Subject Control.” In Standards McIlwaine, I. C. 1998. “Knowledge Classifications, Biblio- for the International Exchange of Bibliographic Information; Pa- graphic Classifications and the Internet” In (Eds.). pers Presented at a Course Held at the School of Library, Ar- (1998). Structures and Relations in Knowledge Organization: chive and Information Studies, University College London, 3-18 Proceedings of the Fifth International ISKO Conference, Lille, August 1990, ed. I. C. McIlwaine. London: Library As- France, August 25-29, 1998, ed. Widad Mustafa El Hadi, sociation, 151-156. J. Maniez, J. and A. S. Pollitt. Advances in Knowledge McIlwaine, I. C. 1993. Guide to the Use of UDC: An Introduc- Organization 6. Würzburg: Ergon, 96-104. tory Guide to the Use and Application of the Universal Decimal McIlwaine, I. C. 1998. “Some Problems of Context and Classification. The Hague: International Federation for Terminology.” Information Studies 4: 195-201. Information & Documentation (FID). McIlwaine, I.C. 1999. “Improving Communication and McIlwaine, I. C. 1993. “A Proposal for the Revision of Classification in the Next Century.” OCLC Newsletter 237: UDC Class 2 Religion with General Observations on 29-31. 576 Knowl. Org. 46(2019)No.8 Obituary. Emeritus Professor Ia McIlwaine: An Appreciation
the Introduction of Greater Facet Analysis into the gion.” In Knowledge Organization for a Global Learning Society: UDC.” Extensions & Corrections to the UDC 15: 31-44. Proceedings of the Ninth International ISKO Conference (Vi- McIlwaine, I. C. 1994. “UDC: The Present State and Future enna, Austria, July 4-7, 2006), ed. G. Budin, C. Swertz and Developments.” International Cataloguing and Bibliographic K. Mitgutsch. Advances in Knowledge Organization 10. Control 23, no. 2: 29-33. Würzburg: Ergon, 323-30. McIlwaine, I. C. 1994. “Africa in the UDC.” African Research McIlwaine, I. C. and N. J. Williamson. 1993. “Future Revi- & Documentation 65: 10-35. sion of UDC: Progress Report on a Feasibility Study for McIlwaine, I. C. 1995. Guide to the Use of UDC. Rev. ed. The Restructuring” Extensions & Corrections to the UDC 15: 11- Hague: FID. 17. McIlwaine, I. C. 1995. “UDC Centenary: The Present State McIlwaine, I. C. and N. J. Williamson. 1994. “A Feasibility and Future Prospects.” Knowledge Organization 22: 64-69. Study on the Restructuring of the Universal Decimal McIlwaine, I. C. 1997. "The Universal Decimal Classifica- Classification into a Full-Faceted Classification Sys- tion: Some Factors Concerning its Origins, Develop- tem.” In Proceedings of the Third International Society for ment, and Influence." Journal of the American Society for In- Knowledge Organization Conference: Knowledge Organization formation Science 48: 331–39. [Also published in Historical and Quality Management, Copenhagen, Denmark, 20-24 Jun Studies in Information Science, ed. M. Buckland & Trudi Bel- 94, ed. Hanne Albrechtsen and Susan Oernager. Ad- lardo Hahn. Information Today: Medford, NJ, 94-106.] vances in Knowledge Organization 4. Frankfurt/Main: McIlwaine, I. C. 1998. “UDC—Into the 21st Century.” In Indeks. 406-13. Globalization of Information: The Networking Information So- McIlwaine, I. C. and N. J. Williamson. 1995. “Restructuring ciety. Proceedings of the 48th FID Conference and Congress, 21- of Class 61–Medical Sciences.” Extensions & Corrections 25 October, 1996, Graz, Austria. The Hague: FID. 92-97. to the UDC 17: 11-66. [Also published in Aslib Proceedings 50, no. 2: 44-48.] McIlwaine, I. C. and N. J. Williamson. 2008. “Medicine and McIlwaine, I. C. 2000. The Universal Decimal Classification: A UDC: The Process of Restructuring.” In Culture and Guide to its Use. The Hague: UDC Consortium. Identity in Knowledge Organization. Proceedings of the Tenth McIlwaine, I. C. 2000. “UDC in the Twenty-First Century.” International ISKO Conference 5-8 August 2008 Montréal, In The Future of Classification, ed. A.R. Maltby & R. Mar- Canada 2008. Advances in Knowledge Organization 11. cella. Aldershot: Gower, 93-104. Würzburg: Ergon, 50-55. McIlwaine, I.C. 2003. “The UDC and the World Wide Web.” In Subject Retrieval in a Network Environment: Proceed- Herculaneum ings of the IFLA Satellite Meeting held in Dublin, OH, 14-16 August 2001, ed. I C McIlwaine. München: K G Saur, McIlwaine, I. C. 1984. “Sir Joseph Banks and the Hercula- 170-76. neum Papyri.” In Proceedings of the XVII International McIlwaine, I. C. 2004. “A Question of Place” In Knowledge Congress of Papyrology, Napoli 19-26 Maggio 1984. Vol. 1. Organization and the Global Information Society: Proceedings Napoli: Centro Internazionale per lo Studio dei Papiri of the Eighth International ISKO Conference London, Eng- Ercolanesi, 197-205. land, July 13-16, 2004, ed. I. C. McIlwaine. Advances in McIlwaine, I. C. 1988. Herculaneum: a Guide to Printed Sources. Knowledge Organization 9. Würzburg: Ergon, 179-85. Volumes 1 and 2. Naples: Bibliopolis for Centro Interna- McIlwaine, I. C. 2007. The Universal Decimal Classification: A zionale per lo Studio dei Papiri Ercolanesi. Guide to its Use. Rev. ed. The Hague: UDC Consortium. McIlwaine, I. C. 1988. “British Interest in the Hercula- McIlwaine, I. C. 2017. “Universal Decimal Classification.” In neum Papyri, 1800-1820.” In Proceedings of the XVIII In- Encyclopedia of Library & Information Sciences, 4th ed., ed. ternational Congress of Papyrology, Athens, 25-31 May 1986. John D. McDonald and Michael Levine-Clark. Boca Ra- Vol. 1. Athens: Greek Papyrological Society, 321-29. ton, FL: CRC Press, 4783-90. McIlwaine, I. C. 1990. “Herculaneum: a Guide to Printed McIlwaine, I.C. and B. Goedegebuure. 1998. “Zukun- Sources: Supplement.” Cronache Ercolanesi 20: 87-128. ftsperspektive der UDK.” In Erschließen, Suchen, Finden: McIlwaine, I. C. 1992. “Davy in Naples: the British View- Vorträge der 19. und 20, hrsg. H.-J. Hermes & J.-H. point.” In Proceedings of the XIX International Congress of Wätjen. Jahrestagungen der Gesellschaft für Klassifika- Papyrology, Cairo, 2-9 Sept 1989. Vol. 1. Cairo: Ain Shams tion: Basel 1995/Freiburg 1996, 137-56. University Center of Papyrological Studies, 107-13. McIlwaine, I. C. and J. H. St. J. McIlwaine. 2002. “Photog- McIlwaine, I. C. 2009. Herculaneum: A Guide to Sources, 1980- raphy – a Proposal.” Extensions & Corrections to the UDC 2007. Naples: Bibliopolis for Centro Internazionale per 24: 62-72. lo Studio dei Papiri Ercolanesi. McIlwaine, I. C. and J. S. Mitchell. 2006. “The New Ecu- menism: Exploration of a DDC/UDC View of Reli- Knowl. Org. 46(2019)No.8 577 Obituary. Emeritus Professor Ia McIlwaine: An Appreciation
University College London School of Librarianship McIlwaine, I. C. and S. R. Ranganathan. 1989. “S. R. Ranganathan: Distinguished Alumnus of University McIlwaine, I. C. 1992. “Ranganathan and University Col- College London.” South Asia Library Group Newsletter 34. lege London.” In S. R. Ranganathan and the West, ed. R. Thorold, I. C. 1965. “The School of Librarianship and Ar- N. Sharma. New Delhi: Sterling, 32-41. chives, University College London.” Library World 67: McIlwaine, I. C. and J. H. McIlwaine. 1987. “Alma Mater 133-35. for Commonwealth Librarians.” COMLA Newsletter 58.
578 Knowl. Org. 46(2019)No.8 H. Moulaison Sandy and A. Dillon. Mapping the KO Community
Mapping the KO Community† Heather Moulaison Sandy* and Andrew Dillon** *University of Missouri, iSchool, 303 Townsend Hall, Columbia, MO 65211, U.S.A.,
Heather Moulaison Sandy is Associate Professor at the iSchool at the University of Missouri. Her research fo- cuses primarily on the intersection of the organization of information and technology and includes the study of issues pertaining to metadata, standards, and scholarly communication. An ardent Francophile, Moulaison Sandy is also interested in international aspects of access to information.
Andrew Dillon is the V. M. Daniel Professor of Information at the School of Information, UT-Austin where he also holds appointments in psychology and info, risk and operations management. He studies the psychological dynamics of information use and design. Dillon is co-editor of the journal Information & Culture and serves on the editorial boards of the Journal of Documentation and Interacting with Computers.
Moulaison Sandy, Heather and Andrew Dillon. 2019. “Mapping the KO Community.” Knowledge Organization 46(8): 578-589. 7 references. DOI:10.5771/0943-7444-2019-8-578.
Abstract: Knowledge organization (KO) is considered a distinctive disciplinary focus of information science, with strong connections to other intellectual domains such as philosophy, computer science, psychology, sociol- ogy, and more. Given its inherent interdisciplinarity, we ask what might a map of the physical, cultural, and intellectual geography of the KO community look like? Who is participating in this discipline’s scholarly discus- sion, and from what locations, both geographically and intellectually? Using the unit of authorship in the journal Knowledge Organization, where is the nexus of KO activity and what patterns of authorship can be identified? Cultural characteristics were applied as a lens to explore who is and is not participating in the international conversation about KO. World Bank GNI per capita estimates were used to compare relative wealth of countries and Hofstede’s Individualism dimension was identified as a way of understanding attributes of countries whose scholars are participating in this dialog. Descriptive statistics were generated through Excel, and data visualiza- tions were rendered through Tableau Public and TagCrowd. The current project offers one method for examin- ing an international and interdisciplinary field of study but also suggests potential for analyzing other interdisciplinary areas within the larger discipline of information science.
Received: 5 September 2019; Revised: 9 September 2019; Accepted: 26 October 2019
Keywords: authors, knowledge organization, articles, research
† Presented at NASKO 2019: Knowledge Organization: Community and Computation, sponsored by ISKO-Canada/United States at Drexel University, Philadelphia, Pennsylvania, USA, June 13-14, 2019.
1.0 Introduction or to position it intellectually both academically and pro- fessionally. Knowledge organization (KO) is sometimes narrowly con- Academic journals provide a forum for the exchange of ceived as a concern of library and information science pro- new knowledge in a discipline and serve as a record of the fessionals, but even a quick examination at the affiliations contributions made to a domain or field across time. As of authors publishing in the field reveals that other intel- such, a scholarly journal serves to validate research, and by lectual domains such as philosophy, computer science, extension, helps to shape the legitimacy of a field of en- business, psychology, linguistics, sociology, and more con- quiry. Long-standing journals in a domain are considered tribute to and find value in its study The subject matter of to provide a measure of prestige for authors as well as an KO embraces fundamental questions of what constitutes identity for a discipline. New areas of enquiry or research knowledge as well as practical concerns of how to repre- involving non-traditional methods often face a challenge sent and enable access for others. Accordingly, it can be gaining a foothold in academia until a suitable peer-re- difficult to characterize and understand the domain of KO viewed outlet such as an academic journal or high prestige Knowl. Org. 46(2019)No.8 579 H. Moulaison Sandy and A. Dillon. Mapping the KO Community conference accepts the work for publication or presenta- 1.2 Metrics to assess countries, comparatively tion. By virtue of this type of gatekeeping role, academic jour- Broad estimates of global expenditure on research sug- nals can provide useful indices of the development of a do- gests where scholarly efforts are most actively pursued, main and the research participants within it. Consequently, and it is perhaps not surprising that in 2017 the US and it is possible to use the back issues of a journal as a test base Europe accounted for over 45% of annual spending on for examining the emergence, duration, and impact of ideas research and development, with China accounting for a within a field, as well as the productivity of key scholars. further 22% (Statista 2019). These proportions correlate Darmani, Dwaikat and Portilla (2013), for example, ana- with the existence and growth of universities globally, lyzed ten years of contributions to the Journal of Creative In- though the US continues to dominate regional presence novation and Management to shed light on how the field of in- within top research university rankings. Domain or disci- novation management is evolving over time and to deter- plinary differences, though more difficult to determine, mine the geographical make up of scholarship in the do- also exist and are likely to reflect national and political em- main. By characterizing author geography, publication phases on research. Chinese universities, for example, are trends, and recurring themes across a decade, they provided becoming highly ranked in engineering and computer sci- evidence of the diminishing occurrence of single-author pa- ence but less so on liberal arts, which remain dominated pers, the recent growth of scholarship from emerging econ- by US and European, particularly British, institutions. omies, and the dominance of leadership as a primary re- Global rankings and expenditures are somewhat limited search emphasis. Similarly, Wiid, du Preez and Wallström measures, and we recognize that scholars can, depending (2012) performed an analysis of Marketing Intelligence and on their circumstances, be mobile, gravitating toward and Planning to identify major author patterns and content trends succeeding at institutions that allow for them to investigate in the field of marketing, highlighting the location of key questions of interest using the methods that are most ap- authors and the productivity of regions in generating new plicable. Further, we must acknowledge that scholarship in scholarship. Such work can prove useful in encouraging a different countries varies in its reward and recognition, and shared perspective and identifying areas of need within a political and economic support from the public and private subject or discipline. The present paper represents an at- sectors. Given the range and the regional differences in tempt at a similar analysis in the domain of KO. support and emphasis for particular research, it is interest- ing to consider where KO scholarship is situated and how 1.1 Efforts to assess the KO community it is distributed and enacted globally. A number of metrics are available to assess cultural dif- Previous studies of KO have addressed questions relating ferences, the best-known being those put forth by Hof- to the field’s geographic reach and intellectual focus. Zhao stede, Hofstede and Minkov (2005). Their metrics, derived and Wei (2017), for example, study collaborations among from large-scale and long-term surveys, outline six dimen- Chinese authors in KO from 1992 through 2016. In exam- sions of culture and profile countries and regions based on ining 1,298 articles with Chinese authors published in Web their scores across these dimensions. As imperfect as these of Science Core Collection KO journals, they find an in- metrics may be, they have become widely used in business crease in collaborations over the period of study, including and research and offer a starting point for comparing cul- in international collaborations (from 50% in1992 to tures internationally. In particular, the individualism vs. 92.53% in 2016). Likewise, Smiraglia (2015) investigates collectivism dimension has the potential to provide insight the field to evaluate the work being done in the area of into the collaborative nature of scholarship and the writing domain analysis, a unique area of study covered in KO. process around the world. We might expect, for example, Beyond KO, scholars in LIS have studied the international that cultures differing on this dimension also manifest dis- contributions to the Journal of the American Society for Infor- tinctive publication styles in terms of single-authored or mation Science and Technology (JASIST) and in the Journal of collaborative articles. Further, we might anticipate that Documentation (He and Spink 2002) over a fifty-year period KO, with its interrogation of knowledge structures and au- at the time when electronic journals were changing the thority might be impacted by cultural distinctions based on scholarly communication landscape. Analyzing first author power distance or uncertainty avoidance. affiliations only, these authors report that international Another metric, put forth by the World Bank, assesses contributions increased over the time of study (1950- relative wealth of a country’s citizens by calculating the gross 1999) for both journals. The extent to which KO mirrors national income (GNI) of the country on a per-capita basis. the broader discipline or represents a distinct area with Limited by virtue of reducing entire populations to a single unique or distinctive scholarly characteristics in its corpus measure of income, these numbers might provide a basis remains an open question. for comparison and, in conjunction with Hofstede et al.’s 580 Knowl. Org. 46(2019)No.8 H. Moulaison Sandy and A. Dillon. Mapping the KO Community dimensions. offer one other gross index to help us better of these could be further analyzed later. Using the individ- understand what we might term the cultural climate of ual author as the primary unit of analysis, each contributor scholarship. to the publication of a scholarly article in Knowledge Organ- ization was identified, and his or her name, institution, 1.3 Mapping the KO community authors school, department, or unit if applicable, the country of the institution, and the total number of co-authors on the For more than forty years, the journal Knowledge Organiza- article were retained in Excel. tion has served as a primary venue for research and dis- Hofstede’s individualism-collectivism dimension was course in the field. As such, the journal contains the richest applied to the data set as a way of understanding relative record of the discipline’s content, contributors, and trends attributes of countries whose scholars are participating in and is explored here to provide us with a database of re- this dialog. World Bank GNI per capita estimates (https:// search activities in the field. Using the unit of “author- data.worldbank.org/indicator/ny.gnp.pcap.pp.cd) in US ship,” we seek to identify what countries appear as a nexus dollars for 2017 were used to compare relative wealth of of KO activity and what patterns of authorship (and co- countries. Because of our assumptions about the mobile authorship) can be found in these data? We wish to char- nature of academics and the observation that English has acterize the KO community of researchers as it has become the lingua franca in scholarly communication, no emerged on empirical grounds to better understand how attempt to understand authors’ country of origin, lan- this area is evolving and how it is positioned intellectually. guages spoken, or educational background was made. De- To begin to explore these questions along with the cul- scriptive statistics were generated through Excel; more tural and disciplinary factors influencing the domain, this complex data visualizations were rendered through Tab- research paper maps the geography of Knowledge Organiza- leau Public and TagCrowd. tion authorship. The current project explores a method for analyzing an international and interdisciplinary field of 3.0 Results and discussion study that we hope might prove useful not just for KO but for other areas of the information discipline in both For this project, 362 scholarly articles, with 632 individual standalone and comparative studies. statements of authors, were coded for analysis and de- scription. In the first instance, we examined publication 2.0 Method rates over time and determined that over the last ten years, there has been almost a doubling of published papers in To assess the question of authorship by nationality based Knowledge Organization (see Figure 1), though this might re- on institutional affiliation, all scholarly articles published in flect exceptional years 2016-2017. Nevertheless, the gen- Knowledge Organization from 2009 to 2018 inclusive were ex- eral trend is positive with increasing number of papers amined. New articles that presented research including re- published in Knowledge Organization over time. search articles and revised conference proceedings were A total of 466 unique authors contributed to the arti- considered scholarly and were retained for analysis. For cles, with the majority (n=384) of authors contributing to this project, scholarly articles retained included articles la- one article, and a minority (n=82) contributing to two or beled “peer reviewed” and research articles that expand on more articles (see Figure 2). What this means for Knowledge peer-reviewed conference proceedings (usually indicated Organization as a scholarly venue is not obvious. This might in the TOC as “Selected Papers from the X Conference” reflect the increasing breadth of new authors publishing – N.B. these tend to be grouped geographically by ISKO in Knowledge Organization or it could be the case of scholars chapter, which affects the mapping of authorship in a way just publishing once here and moving on or not publishing that should be acknowledged. These are nonetheless part further (in the case of students who publish with profes- of the scholarly record produced by Knowledge Organization, sors but then pursue professional careers elsewhere). This so excluding them would be a mistake). Finally, “Reviews is one question that might be usefully pursued over time. of Concepts” in Knowledge Organization were retained. Edi- In terms of individual author productivity, twelve au- torials, features, brief communications, discussions such as thors published four or more scholarly articles over the the “Forum: The Philosophy of Classification,” “Classifi- ten-year period (see Table 1). While traditional author im- cation Research,” “Research Trajectories,” conference re- pact and productivity measures are not the focus of this ports, “ISKO News,” book reviews, introductions to spe- work, it is interesting to note that these twelve individuals’ cial issues, festschrift articles reviewing the life of hon- contributions represent roughly 24% of the journal’s total orees, and reprintings of previously published articles were output. Without comparative data from other fields it is not retained for inclusion. Editorials, book reviews, and re- hard to draw conclusions here but at first glance, this pro- prints of seminal articles were were also excluded, but any portion of contributions from a rather small set of schol- Knowl. Org. 46(2019)No.8 581 H. Moulaison Sandy and A. Dillon. Mapping the KO Community
Figure 1. Scholarly articles appearing in Knowledge Organization by year. Interactive map available online: https://public.tableau.com/pro file/heather8449#!/vizhome/MappingKOauthorship/KOarticlesbyyear.
Figure 2. Number of articles published by each author over the ten-year period. ars might be indicative of an emerging rather than a ma- 1) are generally from the countries with the highest repre- ture field and is likely of some interest to those involved sentation over time, including the United States, Canada, in promotion and tenure discussions. Brazil, and Denmark. Authors were affiliated with institutions located in With authors as the unit of analysis, entries for each thirty-nine countries. See Figure 3 for a breakdown of the author responsible for the scholarly articles studied were number of authors from Algeria to Singapore by year. This coded separately. Figure 4 maps the contributions of these suggests that KO scholarship is indeed global. As ex- authors, by entry for author. Darker blue countries had pected, the most productive scholars shown above (Table higher numbers of total author contributions during the 582 Knowl. Org. 46(2019)No.8 H. Moulaison Sandy and A. Dillon. Mapping the KO Community
Author Country School or department affiliation Articles contributed Birger Hjørland Denmark Department of Information Studies 14 Daniel Martínez-Ávila Brazil Department of Information 12 Claudio Gnoli Italy Library 7 José Augusto Chaves Guimarães Brazil Graduate School of Information Science 7 School of Information Studies, Knowledge Or- Richard P. Smiraglia USA 7 ganization Research Group Elaine Ménard Canada School of Information Studies 6 Joseph T. Tennis USA Information School 6 Margaret E. I. Kipp USA School of Information Studies 6 Melodie J. Fox USA School of information studies 6 Rick Szostak. Canada Department of Economics 6 Fabio Assis Pinho Brazil Department of Information Science 5 Patrick Keilty Canada Faculty of Information 4
Table 1. Individual authors contributing four or more scholarly articles to Knowledge Organization, 2009-2018 and their country and school/de- partment affiliations.
Figure 3. Number of Knowledge Organization authors per year by country. Full visualization can be accessed online: https://public.tab- leau.com/profile/heather8449#!/vizhome/MappingKOauthorship/Authorsperyearbycountry. ten-year period of study with the largest number of schol- (2017), ISKO-Brazil’s 2015 conference (2016), ISKO arly article authors coming from the United States (n=137) Spain-Portugal’s 2015 conference (2016), ISKO-Can- and Brazil (n=105). ada/US’s 2015 conference (2015), ISKO-Brazil’s 2013 The progression over time of international authorship conference (2014), ISKO Spain and Portugal’s 2013 con- can be seen in Figure 5 (interactive version available ference (2014), German ISKO’s 2013 conference (2013), online). Also visible is the publication of the revised pro- ISKO Italy’s 2011 conference (2012), ISKO-France’s 2011 ceedings of the various biennial ISKO chapter meetings conference (2012), and others). The biennial international (featured chapters include ISKO France’s 2017 conference ISKO conference has also been represented. For example, (2017), ISKO-UK’s 2017 conference (2017), ISKO-Bra- the ISKO Conference 2016 (2016) was also featured. zil’s 2017 conference (2017), ISKO-Italy’s 2017 conference Knowl. Org. 46(2019)No.8 583 H. Moulaison Sandy and A. Dillon. Mapping the KO Community
Figure 4. Average number of authors per article by country for 2009-2018 inclusive. An interactive version of this map is available online: https://public.tableau.com/profile/heather8449#!/vizhome/MappingKOauthorship/Authorsbycountry.
Below the national level, we coded authors in terms of in- ences exist. Knowledge Organization has a great deal of interest stitutions, usually universities, and, where provided, with from authors in what Hofstede et al deem more “individu- the academic unit such as school, department, college, etc. alist” cultures, including Canada, the United States, Great Taking these names supplied by authors, a broad overview Britain, and Australia. In fact, Australia, a highly individual- of the disciplinary nature of home units can be generated. istic country, averages one author for paper (N.B., only two Although single instances of affiliations with departments papers with an author from Australia were included in the of archaeology, for example, are not depicted in the word dataset). Farther along the spectrum of the individualism- cloud generated, a sense of the most common depart- collectivism dimension is China, a more collectivist culture ments is available from scanning Figure 6. “Information” in Hofstede’s survey, and indeed Chinese scholars publish is the overarching school/department name, with “library” papers with an average of over three authors. and “computer” perhaps unsurprisingly next in propor- When the average number of authors per article by tion. Interestingly, “communication,” “management,” “en- country is plotted against a country’s individualism-collec- gineering,” “economics,” “business,” and “technology” are tivism dimension score, the trendline reinforces the idea also well represented, creating at least an initial sense that that countries with a higher individualism score like Can- the view of KO as naturally interdisciplinary is supported. ada, Great Britain, the United States, and Australia (aver- Using Hofstede et al.’s (2005) dimension of individual- aging between roughly one and two authors per article) ism-collectivism (the spreadsheet of Hofstede dimensions have fewer average authors per article than more collectiv- used in this project was downloaded from the following ist countries such as Colombia and Pakistan, which average source: https://geerthofstede.com/research-and-vsm/di- four authors from their country per article. See Figure 8. mension-data-matrix/), authors publishing in Knowledge Or- The graph, however, is anything but neat, with the bulk of ganization from countries ranked on this dimension can be the articles having between one and three authors regard- compared to the average number of authors on articles. In less of country of origin. The data in Figure 8 also repre- Figure 7, the darker the color of the country, the higher the sent variations introduced by other cultural dimensions, “individualism” index score. As Hofstede et al. remark (90), but nonetheless, even with the caveats we might place on “The vast majority of people in our world live in societies in the Hofstede model and the limited data set of Knowledge which the interest of the group prevails over the interest of Organization authorship, these trends present an interesting the individual,” but it is clear that significant national differ- lens on authorship and co-authorship. 584 Knowl. Org. 46(2019)No.8 H. Moulaison Sandy and A. Dillon. Mapping the KO Community
Figure 5. Distribution of authorship by country for each year of study. An interactive version of these maps is available online: https://public.tableau.com/profile/heather8449#!/vizhome/MappingKOauthorship/Timelapse2009-2018.
Knowl. Org. 46(2019)No.8 585 H. Moulaison Sandy and A. Dillon. Mapping the KO Community
Figure 6. Word cloud showing alphabetical list of the top fifty of 223 possible words from department or school names with stop words in a number of languages applied (generated using https://tagcrowd.com/).
Figure 7. Individualism and authorship by country. An interactive version of this map is available online: https://public.tableau.com/pro file/heather8449#!/vizhome/MappingKOauthorship/HofstedeIndividualism.
586 Knowl. Org. 46(2019)No.8 H. Moulaison Sandy and A. Dillon. Mapping the KO Community
Figure 8. Hofstede individualism score by country plotted against the average number of authors per country. An interactive version of this figure is available online: https://public.tableau.com/ profile/heather8449#!/vizhome/MAS/ScatterplotIDVxAverageNoAuthors.
Beyond the rate of single or co-authorship, we might ask ical” the next most common title terms. Again, one should if the interests of authors in individualistic and collectivist not draw too firm a conclusion from these trends but they cultures are similar or different? Based on the author’s suggest some differences in emphasis on KO scholarship country of residence, deduplicated lists of the first lines across regions and cultures. of article titles were used to create word clouds for a group Lastly, in considering the geography of contributions of collectivist countries with individualism dimension and relative wealth, Figure 11 presents a map where coun- scores between 18-26, all of which are in East Asia (see tries with larger GNIs are indicated in darker green. Is Figure 9). A second word cloud was created based on titles there a wealth threshold for Knowledge Organization authors? of articles by authors based in the United States (see Fig- Is KO the province of richer or wealthier nations? Contri- ure 10). For both, the term “knowledge” was removed butions seem to be somewhat balanced and there is a range given its frequency in all papers. The East Asian titles rep- of countries on the wealth index participating in KO but resent a smaller set of words (113 possible words) and this is clearly a challenge in all disciplines and one that show greater cohesion, with more words displaying with might be usefully explored further in terms of Knowledge larger font, indicating frequency of use across titles. The Organization’s global growth and reach. presence of “Chinese,” “Mekong,” and “national” suggest perhaps a concern for local initiatives. Interestingly, the 4.0 Conclusion term “organization” does not appear in the East Asian list, which is somewhat surprising given this journal’s coverage. This research presents a first pass at characterizing the In the US titles (a set of 287 possible words), “organiza- international and interdisciplinary community of scholars tion” is predominant, with “analysis,” “domain,” and “eth- publishing in Knowledge Organization. This preliminary anal- Knowl. Org. 46(2019)No.8 587 H. Moulaison Sandy and A. Dillon. Mapping the KO Community
Figure 9. Word cloud showing alphabetical list of the top fifty deduplicated article title words, “knowledge” removed, from countries with individualism indexes 18-26 (i.e., Malaysia, China, Thailand, Singapore, and South Korea) (n=29) (generated using https:// tagcrowd.com/).
Figure 10. Word cloud showing alphabetical list of the top fifty deduplicated article title words, “knowledge” removed, from the United States (individualism score ninety-one) (n=98) (generated using https://tagcrowd.com/).
588 Knowl. Org. 46(2019)No.8 H. Moulaison Sandy and A. Dillon. Mapping the KO Community
Figure 11. GNI of Knowledge Organization authors’ countries by country. An interactive version of this map is available online: https://pub lic.tableau.com/profile/heather8449#!/vizhome/MappingKOauthorship/GNI. ysis suggests four conclusions, with some caveats, as fol- There are clearly several limitations to this work. First, we lows: are using data from only one journal. KO is a field practiced outside of English-speaking areas and thus the contribu- The publication base is growing. Over the last decade tions of non-English language scholars are invisible to this there has been a generally upward growth in the num- project. Further, this is but a preliminary analysis, using a ber of articles published in Knowledge Organization, with limited number of measures for a reduced data set of only the article count doubling from 2009-2018. ten years. While we intend to complete the analysis on the full set of back issues, fewer research papers were published KO research is now a global activity, with published pa- in the early years. Ideally, we would like to compare KO with pers coming not just from the established scholarly other areas within information science to determine if communities in Europe and North America but from Knowledge Organization is unique in its pattern of authorship China and other parts of Asia, the Middle East, South and global activity. Finally, while broad examination of au- America, Africa, and Australia. While the numbers in thor patterns is interesting, it would be instructive to add a some regions are low, there is reason to be optimistic deeper thematic analysis to identify trends in coverage or that KO is establishing itself internationally as a disci- topics that might indicate how Knowledge Organization is pline. evolving over time as well as across regions. It is important to recognize also that direct conversations with authors, par- Authorship patterns indicate that co- or group-author- ticularly those from different regions, would complement ship is routine, but the trend in these numbers suggests this analysis in terms of author motivations, perceived chal- the broad individualist-collectivist distinction of cul- lenges, and sense of intellectual identity in KO. In sum, we tures by Hofstede might help us understand the primary believe there is more work ahead but the early indications differences among regions on this variable. are that such analyses of disciplinary records can prove in- sightful for information scientists. Topical analysis suggests that research in KO may also reflect global cultural differences, particularly on the in- References dividualist-collectivist dimension of Hofstede et al. Our data focused only on two particular regions but is not Darmani, Anna, Nidal Dwaikat and Andres Ramirez Por- exhaustive. tilla. 2013. “Twelve Years of Scholarly Research: Con- Knowl. Org. 46(2019)No.8 589 H. Moulaison Sandy and A. Dillon. Mapping the KO Community
tent and Trend Analysis of the Journal Creativity and Statista. 2019. “Distribution of Research and Develop- Innovation Management.” In 22nd International Confer- ment (R&D) Spending Worldwide from 2017 to 2019, ence for Management of Technology IAMOT, April 14th to by Country/Region.” https://www.statista.com/statis- 18th, 2013, Porto Alegre, Brazil. http://kth.diva-por- tics/732224/worldwide-research-and-development- tal.org/smash/get/diva2:713093/FULLTEXT01.pdf distribution-of-investment/ He, Shaoyi and Amanda Spink. 2002. “A Compairson of Wiid, Ria, Rose du Preez and Åsa Wallström. 2012. "Com- Foreign Authorship Distribution in JASIST and the ing of Age: A 21 Year Analysis of Marketing Intelli- Journal of Documentation.” Journal of the American Society of gence & Planning from 1990 to 2010." Marketing Intelli- Informatino Science and Technology 53: 953-9. gence & Planning 30, no. 1: 4-17. Hofstede, Geert, Gert Jan Hofstede and Michael Minkov. Zhao, Rongying and Xuqiu Wei. 2017. “Collaboration of 2005. Cultures and Organizations: Software of the Mind. New Chinese Scholars in International Articles: A Case York: McGraw-Hill. Study of Knowledge Organization.” Knowledge Organiza- Smiraglia, Richard P. 2015. “Domain Analysis of Domain tion 44: 326-34. Analysis for Knowledge Organization: Observations on an Emergent Methodological Cluster.” Knowledge Or- ganization 42: 602-11.
590 Knowl. Org. 46(2019)No.8 E. Hauser and J. T. Tennis. Episemantics: Aboutness as Aroundness
Episemantics: Aboutness as Aroundness † Elliott Hauser* and Joseph T. Tennis** *University of North Carolina at Chapel Hill, School of Information and Library Science, 100 Manning Hall, Chapel Hill, NC 27599-3360,
Elliott Hauser, MSIS, is a doctoral candidate and at the University of North Carolina at Chapel Hill, where he is a member of the Royster Society of Fellows. His research interests include information organization, data science education, the ethics of algorithms, and sociotechnical methods of systems analysis. His dissertation research on the phenomenon of certainty in information systems was recently selected for the Litwin Books Dissertation Research Award. Elliott has a background in programming, co-founded the online coding education platform Trinket, and teaches programming and information studies topics at the graduate and undergraduate levels.
Joseph T. Tennis is Associate Professor and Associate Dean of Faculty Affairs at the University of Washington Information School, Adjunct Associate Professor in linguistics, and a member of the textual studies, computa- tional linguistics, and museology faculty advisory groups at the University of Washington. He served as President of the International Society for Knowledge Organization from 2014-2018. He is on the Library Quarterly and Knowledge Organization editorial boards and served as a core member of the InterPARES Trust research team from 2005- 2019. Tennis works in classification theory, metadata versioning, ethics of knowledge organization work, descriptive informatics, and authenticity.
Hauser, Elliott and Joseph T. Tennis. 2019. “Episemantics: Aboutness as Aroundness.” Knowledge Organization 46 (8): 590-595. 16 references. DOI:10.5771/0943-7444-2019-8-590.
Abstract: Aboutness ranks amongst our field’s greatest bugbears. What is a work about? How can this be known? This mirrors debates within the philosophy of language, where the concept of representation has similarly evaded satisfactory definition. This paper proposes that we abandon the strong sense of the word aboutness, which seems to promise some inherent relationship between work and subject, or, in philosophical terms, be- tween word and world. Instead, we seek an etymological reset to the older sense of aboutness as “in the vicinity, nearby; in some place or various places nearby; all over a surface.” To distinguish this sense in the context of information studies, we introduce the term episemantics. The authors have each independently applied this term in slightly different contexts and scales (Hauser 2018a; Tennis 2016), and this article presents a unified definition of the term and guidelines for applying it at the scale of both words and works. The resulting weak concept of aboutness is pragmatic, in Star’s sense of a focus on consequences over anteced- ents, while reserving space for the critique and improvement of aboutness determinations within various contexts and research programs. The paper finishes with a discussion of the implication of the concept of episemantics and methodological possibilities it offers for knowledge organization research and practice. We draw inspiration from Melvil Dewey’s use of physical aroundness in his first classification system and ask how aroundness might be more effectively operationalized in digital environments.
Received: 29 September 2019; Revised: 28 October 2019; Accepted: 31 October 2019
Keywords: meaning, aboutness, classification, subject, episemantics
† Presented at NASKO 2019: Knowledge Organization: Community and Computation, sponsored by ISKO-Canada/United States at Drexel University, Philadelphia, Pennsylvania, USA, June 13-14, 2019. We are grateful to an anonymous reviewer for encouraging us to explore the literal aroundness central to early library classification schemes.
1.0 Introduction proaches in relating this idea to the field of KO, and infor- mation studies more broadly. Tennis (2016) proposes epise- This paper discusses and synthesizes two conceptions of mantics as a potential new field of study, analogous to epi- the term episematics developed independently by the au- genetics, just recently made possible due to the advent of thors in prior work. Both conceptions deny that meaning is new technologies and research methods. Hauser (2018a) an inherent property of language, but take distinct ap- asks what it might mean to remove aboutness as a core com- Knowl. Org. 46(2019)No.8 591 E. Hauser and J. T. Tennis. Episemantics: Aboutness as Aroundness ponent of our understanding of information at all. After Classification to literary warrant using the Google Books discussing both proposals, we present a synthesis of each and Hathi Trust corpora (Tennis 2012). Constructing an that connects Tennis’s methodological proposal with episemantic methodology would allow for explicit links, Hauser’s theoretical approach via a shared pragmatism, in revealing how terms were deployed in literature. Star’s sense of “consequences, not antecedents.” The result Essentially, Tennis’s exploratory proposal would allow is discussed in relation to classification theory and particu- subject ontogeny researchers to connect the meaning of larly in light of Melvil Dewey’s pragmatic approach to his subjects to both the use of those subject terms (via large first classification system. Finally, we consider what this scale analysis of cataloging records) and the separate use might mean for organization practices in digital environ- of the same terms outside the context of knowledge or- ments. ganization (via the methods of corpus linguistics). While these methods do not eliminate the methodological con- 2.0 Tennis’s episemantics: Epigenetics for KO cerns Tennis has identified (2016), they represent viable new lines of research with implications for concepts of The idea of episemantics is to account for meaning aboutness and meaning within the LIS context. This would as it changes over time outside of the scheme, and be a nod to studying the semantics and the pragmatics (in relate that to the scheme. Instead of reifying the sub- the linguistic sense) of terms alongside their role in index- ject in the context of the scheme alone, and linking ing and in warrant. Analysis of the “code” of indexing lan- those subjects to a body of documents, episemantics guages in KO could thus be substantially supplemented by would establish models for the investigation of par- examinations of its emergent “expression” within works ticular relationships. These models would be net- and records at scale. We will elaborate on this possibility works of meaning that show how relationships be- below. tween terms are established. 3.0 Hauser’s episemantics: posterior projection Tennis, “Methodological Challenges in Scheme Ver- of meaning sioning and Subject Ontogeny Research,” 578 Losee’s conception of aboutness’s role arises from a Tennis employs an analogy to epigenetics, the study of the category error: while processes’ output is related to effects and behavior of genetic material within living or- both their input and the processes themselves as he ganisms, as opposed to limiting the scope of study to “a” claims, that relationship should not be described as genetic sequence. Epigenetic research has determined that aboutness until episemantic interpretation occurs. the activation and inhibition of specific genes often occurs Following logical empiricism, Losee assumes that in response to environmental or organismal factors in what episemantic interpretation is (or: can be; should be; must be regarded as emergent properties not detectable for science, must be) a transparent process, enabling from a mere sequence of nucleotides. Just as rapid and in- processes’ outputs to be about their inputs. I con- expensive sequencing techniques allowed the relative rates tend that aboutness only obtains in the relationship of expression of genes to be contemplated as a subject of between the interpretation process and the jussive research, thereby enabling a new field, Tennis envisions encoding process. digital methods providing new epistemic access to phe- nomena of deep importance to subject ontogeny research. Hauser, “Information from Jussive Processes,” 303 The challenge that Tennis’s proposed epistemantics ad- dresses is the “location” of meaning in indexing languages Influenced by both the pragmatic philosophy of language in relation to literary warrant. Most indexing languages rely and its continental critics, especially Derrida, Hauser em- on their structure and the intellect of the indexer to trian- phasizes the lack of meaning inherent to inscriptions. For gulate the meaning in indexing terms. Further, meaning Hauser, this is encapsulated in Bowker’s discussion of the can be inferred from the range of materials that are in- jussive. Bowker views “memory practices” in light of the dexed with that term. What has heretofore been lacking is way in which they enact forgetting (Bowker 2006; Hauser the link to the literature except in the rare cases of citations 2018b). For Hauser, this amounts to a proposal to investi- to literature in thesauri (Soergel 1974) and Library of Con- gate technologies of remembering via the techniques of gress Subject Headings (e.g., Library of Congress 2019). How- forgetting they enable. ever, there are no explicit links between these sparse cita- These observations were sparked by a critique of Losee, tions and wider network of literature. who seeks to embed an informative aboutness into a do- Elsewhere, Tennis has presented on the circumstantial main-independent account of information (Losee 1997, evidence relating term appearance in the Dewey Decimal 2012). Losee renders information as the result of pro- 592 Knowl. Org. 46(2019)No.8 E. Hauser and J. T. Tennis. Episemantics: Aboutness as Aroundness cesses and as informative “about” the process and its in- 4.0 Episemantics, recombined puts. This is a powerful approach but problematically em- beds a strong representational aboutness within the foun- Each conception could stand on its own, but we’ve found dation of information. While scientific realists are likely to it generative to consider how the two conceptions might see no problems with such an arrangement, Hauser seeks be recombined. Methodology and theory should ideally re- to preserve the power and expansive domain of Losee’s inforce each other’s strengths to form a coherent whole. work while stripping it of its reliance on scientific realism. Can such a project be accomplished here? Scientific realism is incompatible with many of the do- Tennis’s account is much more deeply embedded within mains we serve, so Hauser tries to preserve as a possible the methodology of classification research, especially sub- viewpoint while avoiding placing it at the core of our dis- ject ontogeny. This depth makes it clear how it might be cipline. applied, but obscures the true power and breadth of the Contra Losee, Hauser locates aboutness as subsequent to idea. Hauser’s approach is more general. This generality the interpretation of inscriptions rather than as inherent to offers greater breadth but is ultimately diffuse and difficult processes. Episemantics is thus the posterior projection of to apply. This section will show how the two approaches meaning (and aboutness) onto inscriptions via interpreta- can be combined to maximize their strengths and mitigate tion. Meaning is always enacted rather than inherent. This each other’s weaknesses. includes both the meaning of information resources and of Tennis’s analogy to epigenetics is apt, and a closer look indexing languages. To revise Losee’s formulation, infor- at the field of epigenetics offers an important template for mation is merely subsequent-to processes; aboutness comes how KO might evolve like traditional genetics when con- afterwards according to Hauser (2018a, 304): “The fronting these ideas. Traditional genetics might simplisti- aboutness relationship consists of and is created by the ep- cally be thought of as a series of sophisticated rules for isemantics of interpretation.” Aboutness is thus not a prop- labeling organisms and groups of organisms. Medical ge- erty but a relation that arises out of interpretive acts. netics uses the possession of genes as, effectively, a cate- Though it is inherently constructivist, Hauser takes pains gorization rule to inform statistical analyses of morbidity to situate scientific realism within this conception. In and mortality (e.g., patients with this gene are X% more Hauser’s reading, Losee’s information from processes and likely to develop heart disease, and live, on average, Y years its embedded aboutness results from a specific account of less than those without). Phylogenetics uses algorithmic the process of interpretation. “Following logical empiricism, measures of similarity to infer ancestral relationships be- Losee assumes that episemantic interpretation is (or: can be; tween species. Each of these approaches contains a step should be; for science, must be) a transparent process, ena- when the object of study is simply labeled genetically, and bling processes’ outputs to be about their inputs” (Hauser from this point on the label is all that is available. This la- 2018a, 303). This framing doesn’t exclude strong represen- beling process is jussive, in Bowker’s sense, and encodes a tationalist conceptions of aboutness but rather de-centers specific disciplinary technique of forgetting. them. They are one amongst many potential instances of Epigenetics represents a deepened interpretation of the creation of meaning. It is this de-centering which ac- DNA sequences by bringing their expression into view. complishes Hauser’s pluralistic goal. As, for better or worse, Traditional genetics was presumed to be a method for a metadiscipline (Bates 1999), we must serve a variety of finding the animating code behind everything but at times fields and individuals, who make meaning in disparate ways. has devolved into a sophisticated mechanism for tagging By identifying these interpretive processes, and cognizing data prior to statistical analyses of co-occurrence patterns. our own, we can better align our activities with the needs of Epigenetics has a claim to this original promise, but must those we serve. do so by abandoning a view of genetic sequences as deter- Losee’s aims, and consequently Hauser’s critique, are of mining the futures of the organisms that possess them in course far broader than classification theory. So how does favor of a more fully contextualized account of how those this work relate to KO? Tennis’s exploratory proposal for genes proliferate and are expressed within an organismal computational analysis of indexing determinations, sup- and ecological context. Wendy Chun has noted the logo- plemented with corpus linguistics becomes an important centrism common to biology and computing technologies way of investigating the enactment of meaning that (Chun 2013). She makes the novel, but convincing, claim Hauser situates at the core of information. Understanding that the kind of logodeterminism represented in works the “posterior projection of meaning” in subject ontogeny like Schrodinger’s “What is Life” was an important precur- thus becomes a project of uncovering evidence of such sor to our understanding of what code is and how com- projections through the analysis of cataloging languages puters work (Chun 2013, Ch. 3). Chun’s analysis suggests within the corpora that surrounded them. a new light within which to view Tennis’s analogy: that ep- isemantics might offer a path, parallel to that of epigenet- Knowl. Org. 46(2019)No.8 593 E. Hauser and J. T. Tennis. Episemantics: Aboutness as Aroundness ics, whereby we gain a greater account of context, and glimpse of what larger scale, comparative “subject phylog- greater explanatory power, by abandoning an outmoded eny” might be. logodeterminism. If we take episemantics seriously, we must revise our Such a potential approach is offered by a pragmatic ac- conception of aboutness. The notion of meaning somehow count of aboutness as aroundness, or “in the vicinity, inhereing in the inscriptions that constitute a language (what nearby.” The occurrence of a gene within a strand of Star might call an “antecedent” view of meaning) has DNA is irrelevant unless the gene is expressed. The import proven philosophically problematic for human languages. of a gene remains unknown precisely “until” we have an Given this difficulty, we suggest abandoning an attempt to account of its expression within an organism, population, clarify or utilize this traditional sense of aboutness for in- or ecological context. Genetic expression is a process of dexing and computer languages. Instead, a turn to pragma- interpretation within a context. Similarly, the possession of tism about meaning and a focus on investigating use, both a term within an indexing language, or applied to a specific within narrow contexts and at scale, offers a viable way for- work, is irrelevant until we know how such a term is used. ward. “Aboutness” in this view need play no larger role than The meaning of a term is impossible to analyze prior to a suggesting that something has been placed near something contextualized account of use. else, as librarians commonly do with cataloged books. The Thus, pragmatism forms a bridge between Tennis and effects of cataloging may be deeply complex, socially em- Hauser’s accounts of episemantics. Pragmatism implicitly bedded, and ethically significant, but the analysis need not animates a good deal of LIS work, and has recently gained include a strong account of aboutness as inherent meaning. traction as a subject of research in its own right (Dousa Rather, we argue, an episemantic approach precludes this. 2009; Buschman 2017; Sundin and Johannisson 2005). Our proposal does not seek to or need to enforce a uni- While competing accounts of pragmatism have been of- form account of aboutness to succeed. Researchers who fered, we prefer Star’s simple and concise definition: a fo- still believe that a strong account of inherent meaning is cus on “consequences, not antecedents” (Star 2015, 133). possible may continue to pursue work in that direction Star here references the words of her mentor Anselm separately. To move forward, we need only agree to pro- Strauss, who in turn was inspired by the work of John ceed with a weak aboutness within the empirically and his- Dewey. This pragmatic ethos unites all three thinkers, even torically oriented study of classification. When we do, Ten- as the meaning of this mantra has evolved. Bowker and nis’s proposal of exactly how this might be studied at scale, Star’s Sorting Things Out would have been far less impactful for both subject ontogeny and the as-yet-unrealized field for our field if it had been subtitled Classification and its An- subject phylogeny, becomes merely a promising suggestion tecedents. A focus on consequences animates both Hauser of many potential ways forward. and Tennis’s approaches. Tennis’s epigenetics analogy shifts focus away from the antecedent, DNA-like indexing 6.0 Aroundness, Dewey, and the digital language to the consequent, RNA-like classification rec- ords and the content of the works they classify. Hauser Although his classification system is often conflated with positions the antecedent inputs of an informative process universalist classification projects, Melvil Dewey himself as ultimately irrelevant to the aboutness of the consequent never considered the “aboutness” of his original classifi- interpretation and enactment of output. cation system to be a specification of the property of the works cataloged and arranged on shelves. In the preface to 5.0 Why does KO need an account of episemantics? the first edition of his classification, it is clear that his focus was primarily on the “effects of placing books near each Episemantics represents an important reminder to avoid other” (Dewey [1876] 1976): viewing meaning as an inherent property of either index- ing terms or abstract concepts. This offers the key meth- In all the work, philosophical theory and accuracy odological benefit of a shared account of both natural and have been made to yield to practical usefulness. The artificial languages in a way that concepts like literary war- impossibility of making a satisfactory classification rant cannot. The materiality of language emphasized by of all knowledge as preserved in books, has been ap- Hauser acts to blur the distinction between natural lan- preciated from the first, and nothing of the kind at- guage, indexing languages, and computer languages. This, tempted. Theoretical harmony and exactness has combined with Tennis’s proposal to look for traces of use been repeatedly sacrificed to the practical require- within all three kinds of languages, presents a new picture ments of the library or to the convenience of the of what classification research might become. In addition department in the college. to strengthening existing techniques such as subject ontog- eny, our recombined concept of episemantics offers a 594 Knowl. Org. 46(2019)No.8 E. Hauser and J. T. Tennis. Episemantics: Aboutness as Aroundness
The effects Dewey considered, of course, were both upon as a future research program. Methodologically, Tennis’s patrons when browsing the shelves, and upon the opera- proposal of utilizing large scale computational linguistics tion and maintenance of the library itself. The fact that as a kind of window into the use and relationship of words subsequent versions of his system and its presentation be- to each other in a corpus would help ground such a project came indelibly associated with universalist classification in the actual use of language rather than encouraging the schemes need not prevent us from returning to it for in- invention and perfection of a crystalline representation of spiration. Dewey’s system, embedded as it was within the the universe of knowledge. Hauser’s exhortation to re- late 19th century library movement’s goals and cultural as- move meaning from classification helps us uncover the sumptions (for more on this, see Miksa 1998), was never- practical effects of classification activities. Dewey’s prag- theless a novel and pragmatic take on how to organize a matism led him to focus on the physical arrangement of newly abundant information resource for optimal use and books. Subsequent modernists sought an ideal, universal management. Viewed in this lens, Dewey’s principle was to space within which to arrange and relate classes to each identify a physical property of information resources, their other. The fragmented space of new digital technologies physical location, and produce a system for manipulating belies either approach. Knowledge is not a set of cartesian this property to balance the needs of library patrons and coordinates, waiting to be arrayed in crystalline perfection. library staff. Though this system contained subject head- There is no reliable experience of physical space to struc- ings, these were merely cogs in an ultimately spatial ma- ture patrons’ encounter with digital resources. How might chine. In our terms, this machine manipulated aroundness we re-envision these organization practices to instead rather than ascribing aboutness. modulate properties that acknowledge the fractured nature Recapitulating this approach with digital resources is of digital encounters but provide flexible structure for non-trivial. Unlike a physical library, the interfaces, se- navigation and exploitation of digital resources? quences, and formats that users access digital information are wildly disparate. To give a simple example, library pa- 7.0 Conclusion trons walk through the front door. Taking this into ac- count, libraries could arrange resources in such a way as to In two separate threads, Tennis and Hauser point to a con- reliably shape these first interactions. Though digital librar- tingent and pragmatic view of aboutness. This leads us to ies still have putative “home” pages, users may land upon reconsider the concept in terms of an earlier meaning, “in practically any part of the site, from practically any other the vicinity, nearby; in some place or various places nearby; digital context. What can serve the function that physical all over a surface.” The vicinity and surface of meaning, proximity did in Dewey’s original system? we have argued, are epistemantically derived—both theo- This, of course, is a question with proliferating answers. retically and methodologically. In a sense, the intractability of organizing the massive Revisiting the early work of Dewey, we uncovered a new amounts of highly specialized knowledge, a task increasingly sense of aroundness, a literal one. Physical location was cen- confronted by Dewey’s successors, encourages the essential- tral to Dewey’s scheme to balance the needs of patrons and ist approach to “aboutness” that we have critiqued. For a library staff. Modernist classification theorists, who Miksa specialist researcher seeking journal articles in her speciality, read as constructing “the universe of knowledge” as their a given resource is either “about” “the desulfurization of domain, still employed an attenuated aroundness in their hot coal gas with regenerable metal” or not. As Miksa notes, schemes relating classes, and thereby subsequently cata- classification theorists who took up the devilish challenge of logued resources, to each other via their physical proximity organizing specialist knowledge, such as Richardson, Bliss, within collections. Dewey’s pragmatism centered around the and Rangagnathan, found themselves increasingly drawn to realization that physical location was the primary “outcome” map a “universe of knowledge,” where every specialist of his classification and the primary tool he had to influence query could have a definite home (Miksa 1998, 56–73 et library operations. seq.). Through the lens with which we have been reading In a digital environment, many possible operationaliza- Dewey’s work, this strikes us as precisely an attempt to pro- tions of aroundness are possible. Commercial information vide an analogy to the physical location that made Dewey’s systems have pioneered many of these, driven by large system work for generalist libraries. A conceptual location scale collection of user data (“Customers who viewed this within the Cartesian space of the universe of knowledge also viewed”). The synthesized conception of episeman- would, modernist classification theory held, allow the pre- tics advanced in this paper is intended to support deep en- cise provision of the right resource for any sufficiently spec- gagements with these new possibilities. We hope that a ified need. pragmatic analysis of the consequences of different man- The task of repeating this process without universaliza- ifestations of aroundness might help provide guidance for tion and its attendant definite aboutness is one we suggest continued innovation in KO. Knowl. Org. 46(2019)No.8 595 E. Hauser and J. T. Tennis. Episemantics: Aboutness as Aroundness
And, of course, episemantics remains an exciting meth- Elisa Cerveira. Advances in Knowledge Organization odological proposal for subject ontogeny research. 16. Würzburg: Ergon, 300-7. Hauser’s theoretical contributions are consonant with the Hauser, Elliott. 2018b. “UNIX Time, UTC, and Datetime: goals of subject ontogeny. If meaning is viewed as enacted Jussivity, Prolepsis, and Incorrigibility in Modern Time- interpretation, examining traces of such interpretation in keeping.” In Proceedings of the ASIST Annual Meeting, corpora and even individual cataloging decisions helps 171-80. Hoboken, NJ: Association for Information Sci- provide insights unavailable by any other means. Amongst ence and Technology. its other goals, we hope that this paper encourages the Library of Congress. 2019. S.v., “Information Organiza- large-scale collaborations needed to understand the com- tion (Authority Record).” sh 99001059. plexity of semantic construction that animates KO activi- Losee, Robert M. 1997. “A Discipline Independent Defi- ties, now and in the past. nition of Information.” Journal of the American Society for Information Science 48: 254-70. References Losee, Robert M. 2012. Information from Processes: About the Nature of Information Creation, Use, and Representation. Ber- Bates, Marcia J. 1999. “The Invisible Substrate of Infor- lin: Springer. mation Science.” Journal of the American Society for Infor- Miksa, Francis L. 1998. The DDC, the Universe of Knowledge, mation Science 50: 1043-50. and the Post-Modern Library. Albany, NY: Forest Press. Bowker, Geoffrey C. 2006. Memory Practices in the Sciences. Soergel, Dagobert. 1974. Indexing Languages and Thesauri: Cambridge, MA: MIT Press. Construction and Maintenance. Los Angeles: Melville. Buschman, John. 2017. “Once More unto the Breach: Star, Susan Leigh. 2015. “Living Grounded Theory: Cog- ‘Overcoming Epistemology’ and Librarianship’s de Facto nitive and Emotional Forms of Pragmatism.” In Bound- Deweyan Pragmatism.” Journal of Documentation 73, no. 2: ary Objects and Beyond: Working with Leigh Star, ed. Geof- 210-23. frey C. Bowker, Stefan Timmermans, Adele E. Clarke, Chun, Wendy Hui Kyong. 2013. Programmed Visions: Soft- and Ellen Balka, 121-42. Cambridge: MIT Press. ware and Memory. Cambridge, MA: MIT Press. Sundin, Olof and Jenny Johannisson. 2005. “Pragmatism, Dewey, Melvil. (1876) 1976. A Classification and Subject Index Neo-Pragmatism and Sociocultural Theory: Communi- for Cataloguing and Arranging the Books and Pamphlets of a cative Participation as a Perspective in LIS.” Journal of Library. Lake Placid, NY: Forest Press Division, Lake Documentation 61: 23-43. Placid Educational Foundation. Tennis, Joseph T. 2012. “Emerging Concepts in Ontogenic Dousa, Thomas M. 2009. “Classical Pragmatism and Its Analysis.” Paper presented at the Research Fair, Univer- Varieties: On a Pluriform Metatheoretical Perspective sity of Washington School of Information. https:// for Knowledge Organization.” In Proceedings from North digital.lib.washington.edu/researchworks/bitstream/ American Symposium on Knowledge Organization. Vol. 2, 1-9. handle/1773/37974/Tennis2012ResearchFairOnto Hauser, Elliott. 2018a. “Information from Jussive Pro- genic.pdf cesses.” In Challenges and Opportunities for Knowledge Organ- Tennis, Joseph T. 2016. “Methodological Challenges in ization in the Digital Age: Proceedings of the Fifteenth Interna- Scheme Versioning and Subject Ontogeny Research.” tional ISKO Conference, ed. Fernanda Ribiero and Maria Knowledge Organization 43: 573-80.
596 Knowl. Org. 46(2019)No.8 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity
The Respective Roles of Intellectual Creativity and Automation in Representing Diversity: Human and Machine Generated Bias † Vanda Broughton University College London, Department of Information Studies, Gower Street, London WC1E 6BT,
Vanda Broughton is Emeritus Professor of library and information studies at University College London. Her principal research interest is in the development of faceted classification, particularly as it affects different disci- plines. She is editor of the second edition of the Bliss Bibliographic Classification (BC2), and an associate editor of the Universal Decimal Classification. In addition to the published volumes of BC2, she is author of several books on knowledge organization systems and numerous articles and conference papers.
Broughton, Vanda. 2019. “The Respective Roles of Intellectual Creativity and Automation in Representing Di- versity: Human and Machine Generated Bias.” Knowledge Organization 46(8): 596-606. 82 references. DOI:10.5771/0943-7444-2019-8-596.
Abstract: The paper traces the development of the discussion around ethical issues in artificial intelligence, and considers the way in which humans have affected the knowledge bases used in machine learning. The phenom- enon of bias or discrimination in machine ethics is seen as inherited from humans, either through the use of biased data or through the semantics inherent in intellectually-built tools sourced by intelligent agents. The kind of biases observed in AI are compared with those identified in the field of knowledge organization, using religious adherents as an example of a community potentially marginalized by bias. A practical demonstration is given of apparent religious prejudice inherited from source material in a large database deployed widely in computational linguistics and automatic indexing. Methods to address the problem of bias are discussed, including the modelling of the moral process on neuroscientific understanding of brain function. The question is posed whether it is possible to model religious belief in a similar way, so that robots of the future may have both an ethical and a religious sense and themselves address the problem of prejudice.
Received: 18 September 2019; Revised: 12 November 2019; Accepted 14 November 2019
Keywords: machine intelligence, bias, human, artificial intelligence, data
† Presented at ISKO-UK 2019: The Human Position in an Artificial World: Creativity, Ethics and AI in Knowledge Organization, at City University, London, UK, July 15-16, 2019.
1.0 What is artificial intelligence? The term is frequently applied to the project of de- veloping systems endowed with the intellectual pro- There are many and varied definitions of artificial intelli- cesses characteristic of humans, such as the ability to gence, and various synonyms for it. Poole et al. (1998, 1) reason, discover meaning, generalize, or learn from note that “the term ‘artificial intelligence’ is a source of past experience. much confusion,” preferring to call it “computational intel- ligence,” although it is likely that artificial intelligence is to- The same article identifies five key aspects of intelligence, day the more widely recognised term. Other names include whether human or machine: learning, reasoning, problem “machine intelligence,” “synthetic intelligence” (Brachmann solving, perception, and language. At the operational level, 2005; Gorg et al. 2014), and “augmented intelligence” (Ojala machine intelligence can take a number of forms: pattern 2018; Albrecht et al. 2015; Hannay 2014) recognition, voice recognition, image (including facial im- The Encyclopedia Britannica (Copeland 2019) defines arti- age) recognition, and machine translation. Copeland’s def- ficial intelligence in the following manner: inition (above) tends towards the narrower field of ma- chine learning: “a form of AI that enables a system to learn Artificial intelligence (AI) [is] the ability of a digital from data rather than through explicit programming … computer or computer-controlled robot to perform “After a model has been trained, it can be used in real time tasks commonly associated with intelligent beings. to learn from data” (Hurwitz and Kirsch 2018, 4-5). Knowl. Org. 46(2019)No.8 597 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity
For the purposes of this paper, artificial intelligence is 3.0 Ethical considerations in knowledge considered mainly within the context of information re- organization trieval, specifically document retrieval, and the ways in which document content can be automatically identified It is now well established that the business of knowledge using intelligent agents. This may involve the construction organization, whether that is classification, indexing, subject of automatic classifiers through machine learning and the representation through headings, or visualization tools, way in which document content is processed. brings with it some ethical concerns. Recently, attention has focussed on fake news, or controversial thinking, such as 2.0 Artificial intelligence as a complement to human holocaust denial, and how such material should be repre- activity sented, but more generally concerns are with the misrepre- sentation or under representation of minority groups, lead- Artificial intelligence has impacted on many areas of human ing to disadvantage and disempowerment. There is now a activity, in part because of the speed with which it can pro- growing body of literature on the ethical theory and philos- cess information in an overloaded world, saving human ef- ophy of KO (Olson 1998; Szostak 2014; Mai 2010, 2013a, fort and apparently offering a more objective way to assess 2013b, 2016), and a substantial number of studies of the and respond to a variety of situations. way in which it can discriminate on the basis of gender Information management is only one of such uses of AI, (Foskett 1971; Marshall 1977; Olson and Ward 1997; Olson where it may promise to solve the perennial problem of or- 2007), sexual orientation (Drabinsky 2013; Fox 2016; How- ganization and retrieval in a situation where there is “too ard and Knowlton 2018), race and ethnicity (Duarte and Be- much to know.” This situation has been acknowledged since larde-Lewis 2015; Adler and Harper 2018), political status the early modern period (Blair 2010), began to be addressed (Lacey 2018), and religion (Broughton 2000; Broughton and by mechanization in the mid-twentieth century, and in the Lomas 2019). twenty-first century prompted numerous studies of the way All of such bias is problematic in a world of increasing in which machines might automatically analyse, categorize, diversity, and the major players in conventional KO are index, and classify documents and other information objects seen to address some of the worst excesses. Factors which through a process generally referred to as automatic exacerbate the bias include: unequal provision either of metadata generation or AMG (Broughton, Palfreyman, and terminology or (in a coded system such as a classification) Wilson 2008; Greenberg et al. 2005). unequal distribution of notation; failure to name at all cer- During that period there had also been a good deal of tain groups or perspectives; and language which has a research into the relative roles of controlled vocabularies strong flavour of one particular favoured perspective or and automatic indexing, generally leading to the conclusion culture. Where culture is a powerful element, as in reli- that a hybrid model offered the best balance between effi- gions, language is a specific problem. ciency and effectiveness; a number of studies demonstrated that the use of a controlled vocabulary improves the perfor- 4.0 Ethical considerations in artificial intelligence mance of the tool or system (Liang et al. 2006; Cheung et al. 2005; Aula and Kaki 2005; Ko et al. 2004). Another major There has been substantial research into the phenomenon theme was the automatic building of classificatory struc- of machine ethics, that is the potential ethical or moral be- tures such as ontologies, independently of humans, from haviour of intelligent agents. It should be carefully differ- text corpora or other sources. The extraction of data from entiated from computer ethics which is concerned with the text continues to be a common means of constructing se- behaviour of humans in the context of computing and in- mantic tools, but, as we may see below, the assumption that formation technology, and with roboethics which refers to terms in text are value free, and mean exactly what they say, ethical behaviour of humans in the design and construc- presents a danger to the usefulness and efficiency of such tion of intelligent machines, and in human-machine inter- exercises. action. Although there are some twentieth-century discus- As AI gains ground as an established tool for processing sions of the possibility of moral—or immoral—actions of and decision making, particularly with respect to personal machines, the field really begins with the 2005 AAAI Sym- data, some questions have been raised as to the acceptability posium on Machine Ethics, where the problem is clearly of AI in this role, the ethics of AI, and the extent to which stated, and named by Anderson et al. (2005): machines can function as intelligent, and as ethical agents. Associated ideas, such as the personhood of robots, and Past research concerning the relationship between whether they can be said to assume responsibility for their technology and ethics has largely focused on respon- actions, have also been considered. sible and irresponsible use of technology by human beings, with a few people being interested in how 598 Knowl. Org. 46(2019)No.8 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity
human beings ought to treat machines. In all cases, sourced by the agent. Additionally, human intervention of- only human beings have engaged in ethical reason- ten supports the machine-learning process through itera- ing. We believe that the time has come for adding an tion with the “teacher,” usually through a technique of ethical dimension to at least some machines. Recog- query-by-example accompanied by feedback to the ma- nition of the ethical ramifications of behavior in- chine. volving machines as well as recent and potential de- Too often the result of this human input is that the ma- velopments in machine autonomy necessitate this. chine inherits the prejudices of the human, so that the bias We explore this dimension through investigation of is hard-wired to the machine (Crawford 2016; Kirchner et what has been called machine ethics. al. 2016; Sears 2018; Kochi 2018). The World Economic Forum (2018, 3) stated: In Anderson et al.’s paper, they consider the implementation of two systems of machine ethics, based on philosophical Designed and used well, machine learning systems principles as displayed in the work of W. D. Ross (theory of can help to eliminate the kind of human bias in de- prima facie duties), and Jeremy Bentham (Utilitarianism), cision-making that society has been working hard to both of which can be expressed as a series of rules. Utilitar- stamp out. However, it is also possible for machine ian ethics and the basis of its decision-making is of particu- learning systems to reinforce systemic bias and dis- lar interest, since in the form of Bentham’s Felicific calculus, or crimination and prevent dignity assurance. Calculus of pleasures (1789), it was designed to be computable, and indeed, one of the themes of the Symposium was the The likelihood of such bias has considerable implications computability of ethics. At the time of Anderson et al.’s re- for human rights, for the proper management of social di- search, the likely guarantee of “good” machine ethics was versity, and for the fair treatment of diverse groups in soci- the imposition of better and more considered rules for the ety. Such inequity is a long-standing problem in conven- machine’s operation, derived from traditional systems of tional information management and has been addressed at ethics and the practice of professional ethicists. As they say length in the research literature of knowledge organization in a subsequent paper (2007, 25): in particular. It seems, however, especially insidious in the machine intelligence context, perhaps because of the expec- Ensuring that a machine with an ethical component tation that higher levels of neutrality and objectivity apply. can function autonomously in the world remains a challenge to researchers in artificial intelligence who 4.2 Bias and discrimination derived from data must further investigate the representation and deter- mination of ethical principles, the incorporation of Much of the literature in this area is centred on machine these ethical principles into a system’s decision proce- decision-making based on demographic data, and the con- dure, ethical decision making with incomplete and un- cern arises from a human rights perspective where some certain knowledge, the explanation for decisions made groups are disadvantaged or marginalized by the way in using ethical principles, and the evaluation of systems which the data is set up (Smith, , Patil and Muñoz 2016; that act based upon ethical principles. Obama White House 2016; World Economic Forum 2018). Generally in these cases, the data is factual and the 4.1 Where machine ethics falls short: bias in decision-making is based on a combination of values in intelligent agents different categories, and the identification and recognition of patterns embedded in data, especially latent associa- The general assessment of machine information pro- tions between one group of attributes and another. cessing and machine decision-making has been that it may Particularly prominent in the discussion of bias in AI is avoid the subjectivity associated with humans. In practice gender discrimination, also a feature of early research into this has turned out not to be the case, since, despite the bias in KO. A recent major study by Criado-Perez (2019) emphasis on machine independence in artificial intelli- reveals that data itself is often biased, because the sample gence, intelligent agents are not created spontaneously, but is in some way flawed. Criado-Perez’s principal concern is require some degree of human participation, and no sys- with gender imbalance, and it is clear that a female per- tem of machine learning can avoid the use of information spective is often omitted, because the data is derived from which has been at some stage processed by humans. The studies that dealt only with males. Criado-Perez provides problem affects equally machine learning where the agent examples of where, for example, diagnostic thresholds has learned from data or a prepared model or training set, based on biomarkers are inaccurate for women, because or in the case of knowledge organization systems, where average figures are based on predominantly male data (as human-constructed vocabularies or ontologies have been in Khamis et al. 2016), since the inclusion of female data Knowl. Org. 46(2019)No.8 599 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity may be as low as 14% of studies in some fields (Pinnow et of automatic classification or term extraction tools, perhaps al. 2014). This may explain many examples of bias de- because work on these originated in the sciences rather than tected in machine intelligence where the agents have been the social sciences and humanities. Automated lexicography trained on datasets that lack comprehensiveness in one or is now a very well-established methodology for building more respects. such semantic tools, whether the lexical data is sourced from other lexical tools such as dictionaries or extracted from text 4.3 Bias and discrimination derived from semantics corpora, and the problem of inherited bias could conse- quently be a serious impediment to both effective retrieval A much-cited paper by Caliskan et al. (2017) establishes that and ethical practice. not only is any incompleteness or skew in the data sample passed on to intelligent systems, but that semantics is also 5.0 Inherited semantic bias in religious terminology: an inheritable factor. Using measurable associations be- the example of WordNet tween pairs of words, Caliskan builds on the work of some prior studies investigating human-like biases in textual cor- WordNet is a vocabulary database, maintained at Princeton pora, particularly that of Greenwald (1998), who studied University, and used extensively as semantic content for all “biases that they consider nearly universal in humans and kinds of automatic indexing and classification tools. It de- about which there is no social concern” (Caliskan, 183). fines itself in the following way (Princeton University 2010): Clear associations between “flowers” and “pleasant,” and “insects” and “unpleasant,” were replicated by Caliskan’s WordNet® is a large lexical database of English. team, as were similar links between “weapon” (unpleasant) Nouns, verbs, adjectives and adverbs are grouped into and “musical instrument” (pleasant), proving the soundness sets of cognitive synonyms (synsets), each expressing of the methodology. Caliskan et al. were also able to repli- a distinct concept. Synsets are interlinked by means of cate more socially significant connections between Euro- conceptual-semantic and lexical relations. The result- pean American names and African American names with ing network of meaningfully related words and con- pleasantness and unpleasantness respectively, a phenome- cepts can be navigated with the browser … Word- non confirmed in practice by Bertrand and Mullainathan Net’s structure makes it a useful tool for computa- (2004) who tested employers’ response to job applications tional linguistics and natural language processing. varying only in the attached European or African sounding names. WordNet is displayed in a thesaurus-like format, although Comparable work was also replicated in the area of gen- the tags it uses are not the conventional ones of information der, associating female names with “family” as opposed to science, but rather offer a more analytical and nuanced range “career,” when compared with male names (Nosek et al. of inter-term relations. Nevertheless it approximates to the 2002a), and the correlation of women with the arts, rather standard tags through its use of the categories hypernym (= than mathematics, or with the sciences (Nosek et al. 2002b). broader term, superordinate class), hyponym (= narrower Studies of inherent discrimination based on religion are term, subordinate class), and synsets (equivalence relation- much less frequent, perhaps because religious affiliation is ships). Other relationships include types (= narrower term much less immediately obvious than gender or ethnicity. generic), instances (= narrower term instantive), and mer- However, Binns (2018, 1) places it on a level with gender onymy (= narrower term partitive). Opening up a “sister and race as a potential factor for discrimination, using the term” reveals terms in the same array, comparable with example of disparate treatment of nationals from Muslim- some kinds of related, or associative terms. This suite of majority countries because of a perceived association of relationships provides the vocabulary with a robust logical Islam with terrorism (2018, 4). Since such examples of re- structure, and verbs as well as nouns are thus organized into ligious prejudice are not uncommon and the problem is hierarchies. Unlike most controlled vocabularies, WordNet one with high public awareness, it is surprising that, to also includes adjectives and adverbs in its database. However date, there is little or no research into bias associated with there is quite limited use of associative term type links, so religious affiliation. the navigation tends to be, on the whole, hierarchical. The existence of such semantic bias has considerable im- WordNet is a good example of a resource that has in- plications for information retrieval because so many auto- herited content. Although initially it was intellectually con- matic classifiers build structures on the back of text corpora structed, it draws on older sources such as thesauri (Baro- on the assumption that these present a neutral and objective cas et al. 2018) that themselves may contain bias, and be- picture of the world. The existence of these biases seem to cause of its widespread use in computational linguistics it be clearly acknowledged in the world of corpus linguistics, passes on that bias. but did not seem to be taken into account at all in the field 600 Knowl. Org. 46(2019)No.8 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity
In some applications, researchers repurpose an exist- many humanities and social science disciplines, particularly ing scheme of classification to define the target varia- those where there is a strong cultural dimension, language ble rather than creating one from scratch. For exam- is a source of some problematic classes and linguistic ex- ple, an object recognition system can be created by pressions. Similarly, the precise analytical structure of training a classifier on ImageNet, a database of images WordNet, based on linguistics principles, while it is highly organized in a hierarchy of concepts. ImageNet’s hi- suitable for the sciences, does not always serve the rather erarchy comes from Wordnet, a database of words, messier humanistic domains nearly as well. Accurate and categories, and the relationships among them. Word- comprehensive category structure is not necessarily to be net’s authors in turn imported the word lists from a found, and in many cases the arrays are incomplete. number of older sources, such as thesauri. As a result, If we consider the standard criticisms of biased religion WordNet (and ImageNet) categories contain numer- classes in standard bibliographic classifications, many of ous outmoded words and associations, such as occu- the same shortcomings are evident in WordNet. For exam- pations that no longer exist and stereotyped gender ple, if we look at the hierarchical display of hyponyms associations. (subordinate classes) under Religion (disregarding the an- notations and further levels of hierarchy) we find: Even a cursory examination of WordNet’s religious cate- gories reveals some very evident examples of bias. As with
Figure 1. Entry for “religion” in WordNet. Knowl. Org. 46(2019)No.8 601 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity
When compared with the standard “big twelve” reli- very appropriate or polite. It seems likely that this rather gions acknowledged by most sources (for example, Hin- uncomfortable content has been imported from much nells 2017; Boyett 2016), WordNet fails to mention Baha’i, older dictionaries without review or amendment. Confucianism, Jainism, Sikhism (other than through its Along with such archaic uses of language, there is also subset Khalsa), Zoroastrianism, and, amazingly, Islam. Ex- a leaning towards a strongly Christian-flavoured under- panding the list to include “full hyponyms” expands the standing of religious terminology, as opposed to a more hierarchy and brings in various Christian denominations, multi-faith approach; examples of this can be seen in the movements within Judaism and Buddhism, and under table below. The terms have been chosen as relatively neu- sects, Anglican High Church, Sunni and Shi’a Islam, the tral ones which occur in a variety of religions, but in defin- Society of Friends or Quakers, Jainism and Hare Krishna. ing the terms or providing synonyms WordNet imposes a Perhaps surprisingly, Scientology appears, but not the broadly Christian interpretation (see figure 2). Mormon Church. In fairness to WordNet, it does contain many religion Needless to say, the omissions, odd associations and specific terms (bhakti, Gemara, hajj, lama, menorah, nir- peculiar language (Hindooism looks very antiquated and vana, shaman, Sufi, synagogue, etc.), but because of the mildly offensive) would be unacceptable in a modern the- mainly hierarchical structure these are not easily accessed saurus or bibliographic classification. Many of the defini- through the parent religion. This positive feature needs tions and verbal qualifications of entries exhibit some odd also to be set against the general Christian tenor of the if not doubtful attitudes, as in the definition: “Hindooism vocabulary and the evident tendency to view religion (a body of religious and philosophical beliefs and cultural through a Christian lens. practices native to India and based on a caste system.” There are also some straightforward factual inaccura- There are differing schools of thought about the caste cies in WordNet, for instance the qualifier of Hinduism system, and whether it arises from socioeconomic rather “(the religion of most people in India, Bangladesh, Sri than religious forces, and this is a very contentious state- Lanka, and Nepal),” whereas the dominant religion in ment. Similarly, Paganism (synonyms: pagan religion, hea- Bangladesh is Sunni Islam, followed by 83.4% of the pop- thenism) is defined as “any of various religions other than ulation (Sawe 2019). Christianity or Judaism or Islamism” which is certainly These significant shortcomings demonstrate a very inaccurate in respect of modern pagans, and potentially considerable bias and a disregard for fairness and sensitiv- offensive to followers of the non-monotheistic faiths. Per- ity towards minority groups. The gravity of bias in Word- haps the worse sufferer is Islam, which is provided with Net is considerably magnified by its widespread use as a the synonyms Muslimism, Mohammedanism, Muham- lexical source for automatic classifiers, which implies that madanism, and Islamism. The Oxford English Dictionary the prejudices will indeed have been inherited and rein- says of Mohammedanism that “its use is now widely seen forced by a great number of other intelligent agents. as depreciatory or offensive,” and none of these terms feel
Figure 2. Synonyms in WordNet. 602 Knowl. Org. 46(2019)No.8 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity
6.0 What solutions exist to the problem of bias? how serious these efforts are. We learn of a Christian robot priest in Wittenberg which radiates light from its hands and As is the case in library and information science, where the pronounces blessings in five languages as part of an exhibi- interests and priorities of the user community demand a tion to celebrate 500 years since the invention of printing privileging of those interests, bias is not always regarded as technology, instrumental in the Reformation and the rise of a bad thing. It is generally agreed that the very fact of a spe- Protestantism (Sherwood 2017). Other cases include a robot cific perspective unwittingly and unavoidably generates bias Buddhist monk in China (Tatlow 2016) which reads scrip- towards the favoured group (such as classification schemes ture and can answer questions, and another in Japan (Field for libraries with specific religious affiliations). Given the im- 2017) which can “chant prayers and tap drums as part of a portance of meeting user expectations and the needs of the funeral ceremony.” user community, bias can be seen as an ethically-neutral phe- There is also a literature in the overlap between religious nomenon. philosophy and AI that considers the nature of the relation- In other cases the investigation of bias is simply a part ships between intelligent agents, humans and the person of of the scholarly study of society and the legitimate search God, typically whether the creation of intelligent agents in for patterns and trends in human cultures. For example, a some sense mirrors the creation of humans (Herzfeld paper by Kozlowski et al. (2019, 38) shows how the ma- 2003), and if the possibilities of transhumanism through the chine analytical technique of word embedding can help to technological alteration of species are realizable (Dumsday reveal changes in social attitudes over time and historic 2017). Vidal (2007, 930) makes a comparison between man’s changes in word meanings. In different situations, the interaction with artificial beings and his interactions with the identification of bias may be the preliminary to addressing gods, asking whether the similarities are not caused by un- it in a social and political context, and is a useful tool in certainty: highlighting social inequalities. In a wider context however, bias should be energetically But it is also true that where interaction is supposed tackled if the system is not to appear as the tool of a par- to exist between the gods and their worshippers, ticular cultural, political, or disciplinary community. Bias there always remains a strong element of uncertainty inherent in data is generally regarded as undesirable and which cannot easily be dismissed concerning the ex- has generated an area of research activity under the general act ontological nature of the hybrid arrangement by heading of machine-learning fairness. Barocas, Hardt, and which the divinity's presence is made manifest. It is Narayanan (2018) provide a broadly-based survey of a precisely the same sort of ontological uncertainty number of problems and potential solutions, based on sta- that one finds expressed in the field of robotics. And tistical adjustment. The book “offers a critical take on cur- this is also why robots both fascinate and worry the rent practice of machine learning as well as proposed tech- general public. nical fixes for achieving fairness.” Mancuhan and Clifton (2014) also propose a statistical 6.2 The moral and religious life of machines solution to bias in data used for automatic financial deci- sion-making, employing Bayesian techniques to identify A pressing question is whether a real sense of moral re- and automatically correct bias. This is incidentally one of sponsibility can be developed in intelligent agents, or, more the few papers to reference religion as an attribute subject fancifully perhaps, a proper religious sense. In human be- to bias, although the authors do not go on to include it in ings, it may seem obvious that ethical decisions differ in their study. some significant respect from other kinds of decisions, and that intellectual reasoning is subordinate to, or at least 6.1 A moral and religious solution strongly influenced by, emotional intelligence. As Liao (2016) says: As with every other area of human life, machine intelligence has impacted religious communities, apart from the general Central area of intellectual inquiry across different philosophical questions of whether robots can act as moral disciplines involves understanding the nature, prac- agents. A number of applications exist which aim to support tice, and reliability of moral judgments. For instance, religious practice, such as the Roman Catholic Confession an issue of perennial interest concerns what moral app (Rau 2011) and Muslim Pro which can tell you prayer judgments are and how moral judgments differ from times and the direction of Mecca in your own town or vil- nonmoral judgments. Moral judgments such as “Tor- lage (Muslim Pro 2019), and attempts have already been ture is wrong” seem different from nonmoral judg- made to use robots in ritual. Most of the literature here is in ments such as “Water is wet.” But how do moral judg- popular journals and the press, so it may be difficult to assess ments differ from nonmoral, but normative judg- Knowl. Org. 46(2019)No.8 603 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity
ments such as “The time on the clock is wrong” or Even if we limit our discussion to computers that “Talking with one’s mouth full is wrong”? are not directly derived from a particular human brain, they will increasingly appear to have their own However, work in neuroscience has questioned whether personalities, evidencing reactions that we can only this distinction between cognitive and emotional aspects label as emotions and articulating their own goals of moral judgements is valid, suggesting instead that all andpurposes. They will appear to have their own free such decisions depend on reasoning through complex cal- will. They will claim to have spiritual experiences. culation rather than a response to stimulus (Woodward And people—those still using carbon‐based neurons 2016). Quartz (2009, 214) states: or otherwise—will believe them.’
Perhaps the most surprising finding to date is that Some specific studies of the relationship between AI and core emotional structures, including the midbrain do- religion include William Sims Bainbridge’s God From the Ma- pamine system and insula, decompose uncertain chine (2006), which investigates the question of whether re- choice contexts along the statistical dimensions that ligious activity might occur spontaneously in machine learn- are the cornerstone of FDT [Financial decision the- ing. Bainbridge and Stark (1987) proposed a general theo- ory] … recent findings suggest that the encoding of retical framework for the scientific study of religion, which value in midbrain dopamine areas might underlie an included, among many other propositions, reasons for the early implicit encoding that is signaled to orbitofrontal emergence of cults and for cult affiliation; individuals char- cortex, where it guides choice. acterized by high levels of education and social isolation, it is suggested, are more likely to participate in cults, a theory Recent studies have shown that it is possible to identify supported in part by practical testing (Bader and Damaris precisely areas of the human brain responsible for social 1996). In 1995 Bainbridge employed the technique of neural and moral behaviour, and to link deficits in brain function network modelling to test his theory, and to identify the rea- to immoral, or amoral, behaviour (Damasio 1994; Shoe- sons for human religious behaviour. The idea of religion, maker 2012). Shoemaker (2012, 807) states: however, is not part of the data supplied (484): “While not denying the possibility of God’s existence, our theory at- The basic limbic emotions are those present in all tempts to explain human religious behavior without assum- mammals emanating from phylogenetically analogous ing the truth of religion. Therefore, there is no axiom as- brain structures collectively called the limbic system serting the existence of the supernatural.” The program at- … These are fear, anger, disgust, sadness, and happi- tempts to formalise communication and models a commu- ness; they function chiefly to promote the survival of nity who seek exchanges with each other in search of the the individual. The moral emotions, the product of basics of existence and of beneficial exchanges leading to the social brain network, arise later in development personal rewards (486 emphasis original): and evolution (Adolphs 2003). They are guilt, shame, embarrassment, jealousy, pride, and altruism; they In the scenario accompanying the program, they are function to regulate social behaviors, often in the called energy, water, food, oxygen, and life. Some people long-term interest of a social group rather than the are producers of one or another of the first four re- short-term interest of the individual person (Adolphs wards, and the simulation models the development 2003). of a little economy based on exchange of these con- sumable rewards. But none of the 24 people can pro- Such work has considerable implications for the develop- vide each other with eternal life. ment of ethically responsible machines, since if the nature of decision-making can be made explicit and the process Bainbridge describes how, in the attempt to find suitable modelled, then it is, at least theoretically, possible to repli- exchange partners for these more intangible rewards, the cate this process in machine decision-making. concept of the supernatural, within the context of a folk A further and more difficult question is whether it is religion, may emerge spontaneously (492): possible to inculcate religious sensibility in machines by a similar methodology. In the speculative Age of Spiritual Ma- In the subculture-evolution model, an intensely in- chines (1999, 6) Kurzweil, technologist and futurist, sug- teracting group of individuals commits itself to the gests that in the distant future this will happen spontane- attainment of rewards, some of which are very dif- ously as the result of technological evolution: ficult or even impossible to obtain. As they exchange rewards among themselves, they also exchange ex- planations about how to get other rewards, and in 604 Knowl. Org. 46(2019)No.8 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity
the attempt to satisfy each other, they magnify ganization from a Social Justice Perspective.” Library slightly their positive evaluation of explanations. Trends 67, no. 1: 52-73. Those explanations that can be evaluated empirically Adolphs, Ralph. 2003. “Cognitive Neuroscience of Hu- will be rejected, leaving the nonempirical (supernat- man Social Behavior.” Nature Neuroscience Reviews 4: 165- ural) explanations that cannot readily be evaluated. 78. Faith will spiral upward, and the group will create a Albrecht, Stefano. V., Andre M. S. Barreto, Darius Brazi- folk religion through a series of thousands of tiny unas, David L. Buckeridge, Heriberto Cuayáhuitl, Nina communication steps. Dethlefs et al. 2015. “Reports of the AAAI 2014 Con- ference Workshops.” AI Magazine 36: 87-98. doi:10.16 In a more developed version of the methodology, Bain- 09/aimag.v36i1.2575 bridge (2006) concludes that this is a natural, and to some Anderson, Michael and Susan Leigh Anderson. 2007. “Ma- extent inevitable, process that he is able to model in vari- chine Ethics: Creating an Ethical Intelligent Agent.” AI ous ways. He posits that it is quite feasible that, primed Magazine 28, no. 4: 15-26. with appropriate theological information, machines might Anderson, Michael, Susan Leigh Anderson and Chris Ar- also “communicate ideas as if they were exchange partners men. 2005. “Towards Machine Ethics: Implementing engaged in theological discussion with each other” (137), Two Action-Based Ethical Theories.” In Machine Ethics: and that “once separated to some degree from external Papers from the AAAI Fall Symposium, ed. Michael Ander- control, the evolving cult develops … the end point of son, Susan Leigh Anderson and Chris Armen. Menlo successful cult evolution is a novel religious culture (139).” Park, CA: AAAI Press, 1-7. https://pdfs.semantic scholar.org/f36b/82a0ee77c9a87f3010f9d0335270d1a2 7.0 Conclusion 4687.pdf?_ga=2.66400620.357820807.1574389742-190 2475365.1566693535 The phenomenon of bias is found to be widespread in ma- Aula, Anne and Mika Kaki. 2005. “Findex: Improving chine intelligence, both in data per se and in semantic con- Search Result Use Through Automatic Filtering Cate- tent derived from text corpora. Most of the studies of ma- gories.” Interacting with Computers 17: 187-206. chine bias have focussed on demographics such as gender, Bader, Chris and Alfred Demaris. 1996. “A Test of the race, and occasionally, social class. Although it is mentioned Stark-Bainbridge Theory of Affiliation with Cults and in a few papers as another potential focus for bias, religious Sects.” Journal for the Scientific Study of Religion 35: 285-303. affiliation has not been investigated in the same way, despite Bainbridge, William Sims. 1995. “Neural Network Models the obvious existence of religious prejudice in society. How- of Religious Belief.” Sociological Perspectives 38: 483-95. ever, examination of the well-established and influential re- Bainbridge, William Sims. 2006. God from the Machine: Arti- source WordNet shows high levels of bias in its structural ficial Intelligence Models of Religious Cognition. Lanham, associations and use of language, almost certainly as a result MD: Altmira. of inheritance from vocabularies used in its original con- Barocas, Solon, Moritz Hardt and Arvind Narayanan. struction. Because of its widespread use in the creation of 2018. “Fairness and Machine Learning: Limitations and search and discovery tools, particularly automatic classifiers, Opportunities.” http://www.fairmlbook.org WordNet is likely to have passed on these prejudices. Bentham, Jeremy. 1789. An Introduction to the Principles of Addressing the problems of bias has mainly concen- Morals and Legislation. Oxford: Clarendon. trated on technical solutions, but recent research in model- Bertrand, Marianne and Sendhil Mullainathan. 2004. “Are ling the ethical behaviour of humans suggests that intelli- Emily and Greg More Employable Than Lakisha and gent agents may be able to develop a sense of fairness and Jamal?” American Economic Review 94: 991-1013. moral responsibility independently of humans, and one not Binns, Reuben. 2018. “Fairness in Machine Learning: Les- assuming pre-programming with specific rules. Several writ- sons from Political Philosophy.” Journal of Machine ers speculate that, in time, a sense of the religious could also Learning Research 81: 1-11. emerge, and can provide a detailed demonstration of how Blair, Ann. 2010. Too Much to Know. New Haven, CT: Yale this might happen using similar neural network methodolo- University Press. gies. In time, robots themselves might be equipped to deal Boyett, Jason. 2016. 12 Major World Religions: The Beliefs, Rit- with the phenomenon of religious prejudice. uals, and Traditions of Humanity's Most Influential Faiths. Berkeley, CA: Zephyros. References Brachman, Ron. 2005.” Getting Back to ‘the Very Idea’.” AI Magazine 26, no. 4: 48-50. Adler, Melissa and Lindsey M. Harper. 2018. “Race and Eth- nicity in Classification Systems: Teaching Knowledge Or- Knowl. Org. 46(2019)No.8 605 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity
Broughton, Vanda. 2000. “A New Classification for the Implicit Cognition: The Implicit Association Test.” Jour- Literature of Religion.” International Cataloguing and Bib- nal of Personality and Social Psychology 74: 1464-80. liographic Control 4: 2-4. Hannay, Timo. 2014. “The Digital Academy and Aug- Broughton, Vanda, Malcolm Palfreyman and Andrew Wil- mented Intelligence.” Information Today 31: 25. son. 2008. “Automatic Metadata Generation for Re- Herzfeld, Noreen. 2003. “Creating in Our Own Image: Ar- source Discovery.” http://www.jisc.ac.uk/media/doc tificial Intelligence and the Image of God.” Zygon 37: uments/programmes/resourcediscovery/metgenre 303-16. port_final_v5.doc Hinnells, John. 2017. A New Handbook of Living Religions. Caliskan, Aylin, Joanna J. Bryson and Arvind Narayanan. Wiley Online. doi:10.1002/9781405166614 2017. “Semantics Derived Automatically from Language Howard, Sara A. and Steven A. Knowlton. 2018. “Brows- Corpora Contain Human-Like Biases.” Science 356, no. ing Through Bias: The Library of Congress Classifica- 6334: 183-6. doi:10.1126/science.aal4230 tion and Subject Headings for African American studies Cheung, C. F., W. B. Lee and Y. Wang. 2005. “A Multi-Facet and LGBTQIA Studies.” Library Trends 67, no. 1: 74-88. Taxonomy System with Applications in Unstructured Hurwitz, Judith and Daniel Kirsch. 2018. Machine Learning Knowledge Management.” Journal of Knowledge Manage- for Dummies. Hoboken, NJ: Wiley. ment 9, no. 6: 76-91. IBM. 2019. “Data Science and Machine Learning.” https:// Copeland, B. J. 2019. “Artificial Intelligence.” Encyclopaedia www.ibm.com/analytics/machine-learning Britannica. Accessed April 14. Khamis, Ramzi Y., Tareq Ammari and Ghada W. Mikhail. Crawford, Kate. 2016. “Artificial Intelligence's White Guy Prob- 2016. “Gender Differences in Coronary Heart Dis- lem.” New York Times. 25 June. ease.” Heart 102: 1142-9. Criado-Perez, Caroline. 2019. Invisible Women: Exposing Kirchner, Julia, Surya Angwin, Jeff Mattu and Lauren Larson. Data Bias in a World Designed for Men. London: Vintage 2016. “Machine Bias: There's Software Used Across the Country Digital. to Predict Future Criminals. And it's Biased Against Blacks.” Damasio, Antonio. 1994. Descartes’ Error: Emotion, Reason ProPublica (blog), 23 May. https://www.propublica.org/ and the Human Brain. New York: Grosset/Putnam. article/machine-bias-risk-assessments-in-criminal-sen Drabinski, Emily. 2013. “Queering the Catalog: Queer tencing Theory and the Politics of Correction.” Library Quar- Ko, Youngjoong, Jinwoo Park and Jungyun Seo. 2004. “Im- terly 83: 94. proving Text Categorization Using the Importance of Duarte, Marisa Elena and Miranda Belarde-Lewis. 2015. Sentences.” Information Processing and Management 40: 65- “Imagining: Creating Spaces for Indigenous Ontolo- 79. gies.” Cataloging and Classification Quarterly 53, nos. 5/6: Kochi, Erica. 2018. “AI is Already Learning How to Discriminate.” 677-702. Quartz (blog), March 15. https://qz.com/author/erica- Dumsday, Travis. 2017. “Transhumanism, Theological An- kochi/ thropology, and Modern Biological Taxonomy.” Zygon: Kozlowski, Austin C., Matt Taddy and James A. Evans. Journal of Religion and Science. 52: 601-22. 2019. “The Geometry of Culture: Analyzing Meaning Field, Matthew. 2017. “This Japanese Robot Can Host Low- through Word Embeddings.” American Sociological Review Cost Buddhist Funerals.” Daily Telegraph, August 24. 84, no. 5: 905-49. doi:10.1177/0003122419877135 Foskett, A. C. 1971. “Misogynists All: A Study in Critical Kurzweil, Ray. 1999. The Age of Spiritual Machines: When Classification.” Library Resources and Technical Services 15: Computers Exceed Human Intelligence. New York: Penguin. 117-21. Lacey, Eve. 2018. “Aliens in the Library: The Classification Fox, Melodie J. 2016. “Legal Discourse’s Epistemic Inter- of Migration.” Knowledge Organization 45: 358-79. play with Sex and Gender Classification in the Dewey Liang, Chun-Yan, Li Guo, Zhao-Jie Xia, Feng-Guang Nie, Decimal Classification System.” Library Trends 64, no. 4: Xiao-Xia Li, Liang Su and Zhang-Yuan Yang. 2006. 687-713. “Dictionary-Based Text Categorization of Chemical Görg, Carsten, Zhicheng Liu and John Stasko. 2014. “Re- Web Pages.” Information Processing and Management 42: flections on the Evolution of the Jigsaw Visual Analytics 1017-29. System.” Information Visualization 13: 336-45. Liao, S. Matthew. 2016. Moral Brains: The Neuroscience of Mo- Greenberg, Jane, Kristina Spurgin and Abe Crystal. 2005. rality. Oxford: Oxford University Press. “Final Report for the AMeGA (Automatic Metadata Mai, Jens-Erik. 2010 “Classification in a Social World: Bias Generation Applications Project).” http://www.loc.gov/ and Trust.” Journal of Documentation 66: 627-42. catdir/bibcontrol/lc_amega_final_report.pdf Mai, Jens-Erik. 2013a. “Ethics, Values, and Morality in Con- Greenwald, Anthony G., Debbie E. McGhee and Jordan L. temporary Library Classifications.” Knowledge Organization K. Schwartz. 1998. “Measuring Individual Differences in 40: 242-53. 606 Knowl. Org. 46(2019)No.8 V. Broughton. The Respective Roles of Intellectual Creativity and Automation in Representing Diversity
Mai, Jens-Erik. 2013b. “Ethics and Epistemology of Clas- Quartz, Steven R. 2009. “Reason, Emotion and Decision- sification.” PowerPoint slides for presentation at II Making: Risk and Reward Computation with Feeling.” Congresso Brasileiro em Organização e Representação Trends in Cognitive Sciences 13: 209-15. do Conhecimento 29 de maio de 2013. http://jenserik- Rau, Andy. 2011. “Should we use a Confession App?” Think mai.info/Papers/2013_EEofClass.pdf Christian (blog), February 24. https://thinkchristian.re- Mai, Jens-Erik. 2016. “Marginalization and Exclusion: Un- framemedia.com/should-we-use-a-confession-app raveling Systemic Bias in Classification.” Knowledge Or- Sawe, Benjamin Elisha. 2017. “Religious Beliefs in Bangla- ganization 43: 324-30. desh.” WorldAtlas, Apr. 25, worldatlas.com/articles/re- Mancuhan, Koray and Chris Clifton. 2014. “Combating ligious-beliefs-in-bangladesh.html Discrimination Using Bayesian Networks.” Artificial In- Sears, Mark. 2018. “AI Bias and the 'People Factor' in AI De- telligence Law 22: 211-38. velopment.” Forbes, November 13. https://www.forbes. Marshall, Joan K. 1977. On Equal Terms: A Thesaurus for Non- com/sites/marksears1/2018/11/13/ai-bias-and-the- Sexist Indexing and Cataloging. New York: Neal Schuman. people-factor-in-ai-development/#555088e59134 Muslim Pro. 2019. “The Most Popular Muslim App!” Sherwood, Harriet. 2017. “Robot Priest Unveiled in Ger- www.muslimpro.com many to Mark 500 Years Since Reformation.” Guardian, Nosek, Brian A., Mahzarin R. Banaji and Anthony G. May 20. https://www.theguardian.com/technology/ Greenwald. 2002a. “Harvesting Implicit Group Atti- 2017/may/30/robot-priest-blessu-2-germany-reforma tudes and Beliefs from a Demonstration Website.” Group tion-exhibition Dynamics 6: 101-15. Shoemaker, William J. 2012. “The Social Brain Network Nosek, Brian A., Mahzarin R. Banaji and Anthony G. and Human Moral Behaviour.” Zygon: Journal of Religion Greenwald. 2002b. “Math = Male, Me = Female, there- and Science 47: 806-20. fore Math ^= Me.” Journal of Personality and Social Psychol- Smith, Megan, Dj Patil and Cecilia Muñoz. 2016. “Big Risks, ogy 83: 44-59. Big Opportunities: The Intersection of Big Data and Civil Obama Whitehouse. Executive Office of the President. 2016. “Big Rights.” Obama White House (blog), May 4. https:// Data: A Report on Algorithmic Systems, Opportunity, and Civil obamawhitehouse.archives.gov/blog/2016/05/04/big- Rights.” https://obamawhitehouse.archives.gov/sites/default/ risks-big-opportunities-intersection-big-data-and-civil files/microsites/ostp/2016_0504_data_discrimination.pdf -rights Ojala, Marydee. 2018. “Digital Ethics in the STM Publish- Stark, Rodney and Willian Sims Bainbridge. 1987. A Theory ing World.” Information Today 35: 10-11. of Religion. New York: David Lang. Olson, Hope A. 1998. “Mapping Beyond Dewey’s Bound- Szostak, Rick. 2014. “Classifying for Social Diversity.” aries: Constructing Classificatory Space for Marginal- Knowledge Organization 41: 160-70. ized Knowledge Domains.” Library Trends 47: 233-54. Tatlow, Didi Kirsten. 2016. “A Robot Monk Captivates Olson, Hope A. 2002. The Power to Name: Locating the Limits China, Mixing Spirituality with Artificial Intelligence.” of Subject Representation in Libraries. Dordrecht: Kluwer. New York Times, April 27. https://www.nytimes.com/ Olson, Hope A. 2007. “How We Construct Subjects: A 2016/04/28/world/asia/china-robot-monk-temple. Feminist Analysis.” Library Trends 56, no. 2: 509-41. html Olson, Hope A. and D. B. Ward. 1997. “Feminist Locales in Vidal, Denis. 2007. Anthropomorphism or Sub-Anthro- Dewey’s Landscape: Mapping a Marginalized Knowledge pomorphism? An Anthropological Approach to Gods Domain.” In Knowledge Organization for Information Retrieval: and Robots. The Journal of the Royal Anthropological Insti- Proceedings of the Sixth International Study Conference on Infor- tute 13: 917-33. mation Research. Hague: International Federation for In- World Economic Forum. Global Future Council on Hu- formation and Documentation, 129-33. man Rights 2016-2018. 2018. “How to Prevent Dis- Pinnow, Ellen, Naomi Herz, Nilsa Loyo-Berrios and criminatory Outcomes in Machine Learning.” http:// Michelle Tarver. 2014 “Enrollment and Monitoring of www3.weforum.org/docs/WEF_40065_White_Paper Women in Post-Approval Studies for Medical Devices _How_to_Prevent_Discriminatory_Outcomes_in_Ma Mandated by the Food and Drug Administration.” Jour- chine_Learning.pdf nal of Women’s Health 23, no. 3. doi:10.1089/jwh. Woodward, James. 2016. “Emotion Versus Cognition in 2013.4343 Moral Decision-Making: A Dubious Dichotomy.” In Poole, David, Alan Mackworth and Randy Goebel. 1998. Moral Brains: The Neuroscience of Morality, ed. S. Matthew Computational Intelligence: A Logical Approach. New York: Liao. Oxford: Oxford University Press, 87-118. doi:10. Oxford University Press. 1093/acprof:oso/9780199357666.003.0004 Princeton University. 2010. “About WordNet.” WordNet. Princeton University. https://wordnet.princeton.edu/ Knowl. Org. 46(2019)No.8 607 Shu-Jiun Chen. Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China
Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China † Shu-Jiun Chen Institute of History and Philology, Academia Sinica, Taipei 11529, Taiwan,
Shu-Jiun Chen is an assistant research fellow at the Institute of History and Philology, Academia Sinica, and also the executive secretary of the Academia Sinica Center for Digital Cultures (ASCDC). She received her PhD degree in library and information science from the National Taiwan University in 2012. Her research interests include digital libraries, metadata, knowledge organization, linked data and digital humanities. Dr. Chen initiated the Chinese-language Art & Architecture Thesaurus (AAT) research project with the Getty Research Institute in 2008, and she is also the principal investigator of the Linked Open Data Lab in ASCDC.
Chen, Shu-Jiun. 2019. “Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China.” Knowledge Organization 46(8): 607-614. 13 references. DOI:10.5771/0943-7444-2019-8-607.
Abstract: The study uses the Database of Names and Biographies (DNB) as an example to explore how in the transformation of original data into linked data, semantic enrichment can enhance engagement in digital human- ities. In the preliminary results, we have defined instance-based and schema-based categories of semantic enrichment. In the instance-based category, in which enrichment occurs by enhancing the content of entities, we further determined three types, including: 1) enriching the entities by linking to diverse external resources in order to provide additional data of multiple perspectives; 2) enriching the entities with missing data, which is needed to satisfy the semantic queries; and, 3) providing the entities with access to an extended knowledge base. In the schema-based categories that enrichment occurs by enhancing the relations between the properties, we have identified two types, in- cluding: 1) enriching the properties by defining the hierarchical relations between properties; and, 2) specifying properties’ domain and range for data reasoning. In addition, the study implements the LOD dataset in a digital humanities platform to demonstrate how instances and entities can be applied in the full texts where the relationship between entities are highlighted in order to bring scholars more semantic details of the texts.
Received: 30 September 2019; Revised: 14 November 2019; Accepted: 15 November 2019
Keywords: open data, linked open data (LOD), property, Database of Names and Biographies (DNB), information, semantic
† Presented at ISKO-UK 2019: The Human Position in an Artificial World: Creativity, Ethics and AI in Knowledge Organization, at City University, London, UK, July 15-16, 2019. The author would like to thank Dr. Cheng-yun Liu for his helpful advice on social history of the Ming and Qing Dynasties in this Paper, and Mr. Lu-Yen Lu, Ms. En-Ting Yu and Dr. Hsiang-An Wang for their assistance throughout the project. The work was supported by the Academia Sinica Center for Digital Cultures (Grant No. ASCDC-107-13), and the Ministry of Science and Technology (Grant No. MOST 106-2410-H-001-025).
1.0 Introduction porting historians’ research. Each metadata record has in- formation including name, alternative name, dates of Semantic relations between entities are the basic units of birth/death, native place, biographical data, work experi- knowledge organization (Green 2001; Stock 2010). This ences, related persons, specialties, academic background, paper discusses the issue of semantic enrichment in linked job titles and references. Semantic enrichment is one of open data (LOD) research. The study will use the Database the current LOD project’s core research streams, devel- of Names and Biographies (DNB) as an example to ex- oped by collaborating with historians for converting the plore how, in the transformation of original data into DNB from legacy databases to linked data format for the linked data, semantic enrichment can facilitate research in- purpose of digital humanities. The study has deployed the quiries and enhance engagement in digital humanities. methods of data modeling, data reconciliation and data en- Hosted by the Institute of History and Philology, Aca- richment during transforming the legacy metadata records demia Sinica (Taiwan), the DNB contains 35,666 records into LOD in order to add value that is structured in a ma- of Chinese historical persons who are cultural and socio- chine-processable format and gives more meaning to the political elites in late imperial China (1368-1911), extracted dataset (Hyvönen 2018; Van Hooland and Verborgh 2014; from various historical archives for the purpose of sup- Zeng 2019). The DNB data model is composed of five 608 Knowl. Org. 46(2019)No.8 Shu-Jiun Chen. Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China core classes (i.e., agent, event, place, object and time clas- AAT (Art & Architecture Thesaurus) concept of “callig- ses), sixty-seven properties from sixteen semantic vocabu- raphy” is directly linked with the agent by the specialty laries (i.e., bio, dbpedia, dcterms, gvp, leo, owl, rdfs, property (dbpedia-owl:specialty). In the original DNB schema.org, skos, etc.) and reuses three external resources metadata, a specialty of Tseng is presented in literal (i.e., AAT, VIAF, TGAZ). The LOD dataset contains as”書法” (calligraphy) in Chinese characters. Since the more than two million triples and is available to query AAT is a multilingual thesaurus including English, Chi- linked data with SPARQL from the LODLab of Academia nese, German, Dutch and more (Harpring 2018), the link Sinica (http://data.ascdc.tw/en/sparql.php). The follow- to the AAT term for “calligraphy” (AAT 300053162) can ing presents the preliminary results of the study, which we enrich the agent entity in the related property with infor- have defined instance-based and schema-based categories mation on the same concept and its definition in other lan- of semantic enrichment. guages. This could enhance understandability for users who do not know the Chinese language. 2.0 Instance-based semantic enrichment 2.2 Enriching the entities with missing data which Instance-based, or entity, enrichment, in the context of is needed to satisfy the semantic queries LOD is a process to endow the original data with other supplementary or supported knowledge, which is beyond As to the information on the native place of a person, since the original data content, from external resources. This is its value shows the historical administrative name of a place done to extend the perspective of the data itself and also and is not entirely equivalent with the current place name, integrate heterogeneous resources into a more complete we, therefore, added a new contextual entity for a place for and sophisticated knowledge. By enriching the entities in each agent when the referred agent has information in the the LOD-based dataset, it not only enlarges the knowledge field of “native place.” The method of adding a place entity base itself but also inspires new perspectives for further is not only to describe the historical name but also to be en- research and interpretation of the study results. There are riched by linking to the linked data of the Chinese historical three approaches identified in the study to achieve entity place name in the Temporal Gazetteer (TGAZ) developed enrichment as follows. by Harvard University, which defines the geographical range and temporal information of the related historical place 2.1 Enriching the entities by linking to diverse, name. Such a place entity is linked to the agent entity of this cross-domain external resources in order to study by the native-place property (ascdc:nativePlace) and provide additional data of multiple also carries the label (rdfs:label) and the link to external re- perspectives sources by the same-as property (owl:sameAs) (shown in figure 1). In the DNB dataset, which is focused on the information Another reason to reuse the terms of Chinese historical of Chinese historical figures, the agent entity is linked to place names from TGAZ is that on the SPARQL (SPARQL the cross-domain external resources. For instance, in the Protocol and RDF Query Language) interface for DNB da- agent entity of Tseng Guo-fan (曾國藩, 1811-1872), a fa- taset, the study applied Chinese historical Geographic Infor- mous literati and high minister of the late Qing period, the mation System (GIS) maps for showing an agent’s native
Figure 1. Enriching the entities with missing data, which is needed to satisfy the semantic queries. Knowl. Org. 46(2019)No.8 609 Shu-Jiun Chen. Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China place in certain examples of SPARQL query. For instance, trolled vocabulary developed by the study for defining the when one queries “which provenances did the Qing agents Chinese historical periods and yearly times of all dynasties ranked as jinshi (進士, “presented scholar,” the highest de- in China (see Figure 2). The application of such term lists gree of the Chinese Imperial civil service examinations) is based on the “time ontology,” which is developed by the come from?” Since the current Google Map as open source study and based on W3C’s time ontology in OWL (Cox shows only the global map in modern political and adminis- and Little 2017). trative boundaries, it is quite different from the historical For example, the DNB person agent Ding Bao-zhen map. The location of a city from hundreds of years ago (丁寶楨) was appointed as Governor of the Sichuan Prov- might also be different from its current geographic situation. ince (四川總督) between 1881 and 1886. In the data Since each term in the Temporal Gazetteer contains the lon- model design of DNB, a person agent’s official career in gitude and latitude information of a historical place name in the government is expressed as entities of official service different dynasties or periods, the application of TGAZ is (ascdc:OfficialService). To describe the beginning date and reasonable for displaying the related geographic information end date of Ding’s appointment, this entity for official ser- in the historical GIS-maps. vice is linked to an entity of temporal period (time:Tem- poralEntity) by at-time property (leo:atTime), which fur- 2.3 Providing the entities with extension to a ther describes the beginning and end date of the related knowledge base period for official service in the historical Chinese era year, represented by the “time:instant” entity. To extend the The extension of data content in the entities to a knowledge base for that temporal information, the study’s knowledge base is regarded in this study as one part of time ontology is applied to describe the hierarchical tem- data enrichment. It could not only enable the entities with poral details of the instances as the “7th year of Guangxu” more detailed information from the external resources but (光緒七年, 1881) and “12th year of Guangxu” (光緒12年, also make linkage of the original entity to an entire 1886) and linked with the entity for era name (光緒), em- knowledge base of a certain subject field, which can in- peror name (光緒皇帝) and dynasty name (清朝) in spire new perspectives for further research and interpreta- “time:propertInterval” by the properties “time:interval- tion of the study results. For instance, the study makes ex- During” and “time:inside”(see Figure 3). tension of entities to the time ontology, which is one part After extending the entity with the temporal terms to of the knowledge base. In the current LOD-based dataset describe the beginning and ending year of a certain period of the DNB, the entities relating to the temporal infor- of an official position, the information of these men- mation on the official career of a person is particularly ex- tioned historical years is expressed as entities of thing and tended to the “term lists of the time and periods,” a con- can further be linked to the era names, emperor names and
Figure 2. Extension of DNB entities with the study’s time ontology, viewed with classes. 610 Knowl. Org. 46(2019)No.8 Shu-Jiun Chen. Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China the dynasty names in Chinese history. Therefore, a hierar- property (agrelon:hasTeacher), while the relationships of chically structured knowledge base, which is focused on agent A to his grandparent, agent C, is expressed as the has- Chinese historical temporal names, is entirely integrated grandparent property (agrelon:hasGrandparent). into the dataset and enriches the data content of each re- In fact, the personal relation between agents is a mutual- lated person agent. or hierarchical-expressible relationship. If A is stu- dent of B, then the B should be the teacher of A. In the 3.0 Schema-based semantic enrichment semantic data model design, such mutual or hierarchical re- lation can be defined by enriching the definition of the re- Schema- or property-based enrichment in LOD is a pro- lated properties. In the current world of semantic web, cess to enable hierarchical or associative meaning within RDFS and OWL are the two types of data vocabularies, pairs of property, which could create a relationship be- which are mostly applied to enrich the relations between tween related entities and also enable a meaningful and ef- properties as the abovementioned cases and also to enhance ficient data query in a hierarchical semantic structure. the efficiency by data reasoning (Polleres, Axel, et al. 2013). In particular, the subproperty-of property (rdfs:subpoper- 3.1 Enriching the properties by defining the tyOf) can be used to mark the hierarchical relation between hierarchical and associative relations between properties, while the inverse-of property (owl:inverseOf) is properties suitable to describe the mutually affected relations between properties on the same level (see Table 1). To a certain extent, the applicability of enriching the prop- Taking the agent Zeng Guo-fan (曾國藩, DNB: erties in the LOD datasets depends on whether a hierar- NO000000058 ) in the DNB datasets as an example, the fig- chical or associative meaning exists between different data ure is linked to the agent Li Hong-zhang (李鴻章, DNB: elements of the original metadata. In the data element of NO000002242) by the has-student property (“agrelon:has- the current DNB datasets, such related meaning is especially Student). Since the relationship between a teacher and stu- found in the data element as the “personal relations” dent is mutually referred, if the inverse-of property (owl:in- (人物關係) of an agent, in which a person’s connections to verseOf) is reused and enriches the definition of the has- another agent is linked and expressed by reusing suitable student property (agrelon:hasStudent), the reverse relation properties. As an example, the semantic relationships of an expressed as has-teacher property (agrelon:hasTeacher) agent A to his teacher, agent B, is defined as the has-teacher would also be findable by data reasoning.
Figure 3. Extension of the entity with knowledge-based external resource (Example: Extension of the entity for official service with hierar- chical information on the beginning and end date of Ding Bao-zhen’s appointment as Governor in Sichuan between 1881 and 1886).
Properties Domain Range Function of property enriching 1 rdfs:subPropertyOf Property Property Hierarchical relation 2 owl:inverseOf Property Property Mutual relation
Table 1. Enrich the properties by using rdfs:subpopertyOf and owl:inverseOf. Knowl. Org. 46(2019)No.8 611 Shu-Jiun Chen. Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China
With the same example of Zeng Guo-fan, the figure is standard by the semantic data model design and also de- further linked to the agent Yuan Bingzhen (袁秉楨, DNB: fines which data context that a property could be reused in NO000012514) by the has-child-in-law property (agre- the data structure. lon:hasChildInLaw) and to Zeng Guangquan (曾廣銓, In the semantic data model design, the domain of a prop- DNB: NO000008193) by the has-grandchild property erty is always the instance of an entity (or a class), which (agrelon:hasGrandchild). Since those different types of re- defines the subject of an information. As in the abovemen- lations to a person can be clustered in a broader range of tioned example, a property as “bio:father” in the DNB is properties such as relatives, a hierarchical structure of prop- applied to describe the information on the father of a per- erty can be hence defined by using the sub-property-of son. The domain of this property is defined as “foaf:Agent,” property (rdfs:subPropertyOf) to structure the has-child-in- which means this property is only suitable to be applied by law property (agrelon:hasChildInLaw) and the has-grand- an entity of the agent. In the machine-processable form, it child property (agrelon:hasGrandchild) both under the could be formulated as “rdfs:domain foaf:Agent.” However, property as has-relative (agrelon:hasRelative). the range of a property could be expressed as a different data type, such as the instance of an entity (or a class), literal 3.2 Specifying properties’ domain and range for information, date decimal numbers of measurement or data reasoning quantity of item. Again, as in the example (Table 2), the specification also defines the range of “bio:father” as In the semantic web, each data can be expressed as a triple, “foaf:Agent,” which means that object information should which is composed of subject, property and object. From also be an instance of agent entity. In the machine-processa- them, the major function of a property is to enable the ble form, it could be formulated as “rdfs:range foaf:Agent.” entities of information (subject and object) with a seman- tic relation, which could enable the data query in a logical, 4.0 Applications of LOD for digital humanities machine-processable and machine-understandable way. In other words, the property plays a role as the bridge to con- The input of the LOD-based dataset of the Database of nect the subject with object and thus construct complete, the Names and Biographies (DNB) into the system of the meaningful information in the data. However, the use of a Digital Humanities Research Platform (DHRP), devel- certain property is not arbitrary. Each property has its own oped by the Academia Sinica Center for Digital Cultures definition to which condition or restriction can be applied (ASCDC), is a linked data application to enhance the reus- to link the subject- and object-entities. ability of the DNB data. Additionally, the study demon- In the current data model design for the “Database of strates the possibility of applying a LOD-based dataset to Names and Biographies” (DNB), sixty-seven properties enlarge the research scope of the scholars in digital hu- from sixteen vocabularies are reused to describe the bio- manities and to integrate into digital research tools- using graphic information on Chinese historical figures and re- examples in the DHRP. lations between figures. The information on the domain The DHRP is an open, cloud-based text repository to and range of all reused properties is described in the spec- enhance the research of digital humanities, which is devel- ification of DNB ontology, seen in the selected examples oped as a platform for online services based on the needs in Table 2. Such specification can be used as referential of scholars. The platform is equipped with different digital tools for text and visual analytics, such as text annotation,
bio:father URI http://purl.org/vocab/bio/0.1/father Label Father Type Property Comment To describe information on father of a DNB Agent Entity Domain foaf:Agent Range foaf:Agent Quantification 0-1 Data type Concept/ASCDC Examples Liu yun father Liu Ton-hsung
Table 2. Specifying domain and range of the father property (bio:father) in DNB. 612 Knowl. Org. 46(2019)No.8 Shu-Jiun Chen. Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China text similarity comparison, N-gram analysis, historical spa- gray background (Figure 4). The type of semantic relation- tiotemporal visualization or social network analysis. The ships (properties) between different agents will be repre- digital content of the DHRP is currently uploaded with sented in a dotted line, which link the subject entity with texts from rare Chinese books for a total of more than 220 its related object entity. million words. In the current stage, the DNB dataset is al- When moving the mouse cursor onto the subject agent ready in the test version of the DHRP-platform. In partic- in the text, the platform will automatically present the re- ular, we use the DNB’s data on properties of the person’s lated agents of person names in green with the type of relations, specialty and native place as practical cases of semantic relation. Further, clicking on those names will di- studies to map the text passages in the Qing Shilu (the Ver- rect to the website of the LOD-datasets, showing the data itable Records of the Qing Dynasty/清實錄) and to content of the related records of the persons (Figure 5). demonstrate the semantic relations between different per- For further presentation of the related data in DHRP- son agents or agent entities with place or concept entities platform by using the tools for data visualization, the func- in the historiographical works of the DHRP-platform. In tion of social network analysis (SNA) is integrated into the total, more than 20,000 named-entities from the 93,431 system to show the matched persons in Qing Shilu based text passages in the Qing Shilu are matched with the in- on the DNB dataset (Figure 6). stances of entities in the DNB. In the original DHRP-platform, the scholars could only In the DHRP-platform, the mapped DNB personal execute the text retrieval and comparison based on the lit- names in the Qing Shilu will be marked up in different col- eral context uploaded in the system. Users could not find ors according to their types in the data unit of a triple in out detailed information or definitions of the retrieved the DNB dataset For instance, the person’s name belong- words or text, since the context was not linked to the ex- ing to the subject in a triple will be highlighted in blue, ternal resources by a sematic method. After uploading the while names of an object in a triple will be shown upon a LOD-based dataset in the originally text-based DHRP-
Figure 4. Revealing the relations between a matched person’s name from DNB in Qing Shilu on the DHRP platform (Example: Teacher- student Relationship between Zeng Guo-fan (曾國藩) andd Li Hong-zhang (李鴻章) are shown with a type of sematic relation (has- Student) in red). Knowl. Org. 46(2019)No.8 613 Shu-Jiun Chen. Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China
Figure 5. Linking the matched person’s name in Qing Shilu to the external resource in the DNB (Example: Moving the cursor onto the person name of Zeng Guo-fan (曾國藩) and showing his related persons retrieved in Qing Shiluo; clicking one of the names as Zeng Guo- chuan (曾國荃) and linking to the equivalent resource in the DNB).
Figure 6. SNA-analysis showing the relation types between matched persons in a passage of Qing Shilu in different forms of data visualization. (1) SNA by e-chart; (2) SNA by D3.js data visualization.
614 Knowl. Org. 46(2019)No.8 Shu-Jiun Chen. Semantic Enrichment of Linked Personal Authority Data: A Case Study of Elites in Late Imperial China platform, a detailed definition of the matched text (for ex- Joint Meetings, ed. Shih-Lung Shaw, Ta-Chien Chan and ample, the further biographical information of a person) Ling-Jyh Chen. Piscataway, NJ: IEEE. doi:10.23919/ can be shown in DHRP by reusing the related data in DNB PNC.2018.8579460 dataset. This is accomplished through the named-entity Hyvönen, Eero. 2018. “Cultural Heritage Linked Data on recognition, which is a mapping procedure of the terms in the Semantic Web: Three Case Studies Using the Sampo DNB with the text in the platform. The semantic relations Model.” In Datu ireki estekatuak eta informazioaren ku- of a person to another person, place or concept will also deaketa integrala ondare-erakundeetan: VIII. Arte Garai- be notified by endowing the type of relations (properties) kideko Dokumentazio Zentroen Topaketak. Vitoria-Gasteiz, in the platform. These could extend not only the Spain: Artium, 19-20. knowledge base of a scholar but might also offer other rel- Manguinhas, Hugo, ed. 2016. “Europeana Semantic En- evant information or inspire research angles, which one richment Framework: Documentation,” Contributors: might not take notice when retrieving the results merely in Antoine Isaac, Valentine Charles, Yorgos Mamakis, Juli- a literal context. ane Stiller. Version: 17th November 2016 (updated 2017, 2018). https://docs.google.com/document/d/1J References: vjrWMTpMIH7WnuieNqcT0zpJAXUPo6x4uMBj1pE x0Y/edit# Baca, Murtha and Melissa Gill. 2015. “Encoding Multilin- Polleres, Axel, Aidan Hogan, Renaud Delbru and Jürgen gual Knowledge Systems in the Digital Age: The Getty Umbrich. 2013. “RDFS and OWL Reasoning for Linked Vocabularies.” Knowledge Organization 42: 232-43. Data.” In Reasoning Web 2013: Reasoning Web. Semantic Tech- Bischof, Stefan. 2017. “Complementary Methods for the nologies for Intelligent Data Access, ed. Sebastian Rudolph, Enrichment of Linked Data.” PhD diss., Technische Georg Gottlob, Horrocks Ian and Frank van Harmelen. Universität Wien. https://aic.ai.wu.ac.at/~polleres/su Lecture Notes in Computer Science 8067. Berlin: pervised_theses/Stefan_Bischof_Dissertation_2017.pdf Springer, 91-149. Cox, Simon and Chris Little. 2017. “Time Ontology in OWL.” Stock, Wolfgang G. 2010. “Concepts and Semantic Rela- Retrieved from https://www.w3.org/TR/owl-time/ tions in Information Science.” Journal of the American So- Ding, Li, Joshua Shinavier, Zhenning Shangguan and Deb- ciety for Information Science and Technology 61: 1951-69. orah L. McGuinness. 2010. “SameAs Networks and Be- Subhashree, S., Rajeev Irny and P. Sreenivasa Kumar. 2018. yond: Analyzing Deployment Status and Implications “Review of Approaches for Linked Data Ontology En- of OWL:sameAs in Linked Data.” In The Semantic Web: richment.” In Distributed Computing and Internet Technology: ISWC 2010, ed. Peter F. Patel-Schneider, Yue Pan, Pas- 14th International Conference, ICDCIT 2018, Bhubaneswar, cal Hitzler, Peter Mika, Lei Zhang.Jeff Z. Pan, Ian Hor- India, January 11–13, 2018, Proceedings, ed. Atul Negi, Raj rocksand Birte Glimm. Lecture Notes in Computer Sci- Bhatnagar and Laxmi Parida. Lecture Notes in Com- ence 6496. Berlin: Springer, 145-60. puter Science 10722. Cham: Springer, 27-49. Green, Rebecca. 2001. “Relationships in the Organization Van Hooland, Seth and Ruben Verborgh. 2014. Linked of Knowledge: An Overview.” In Relationships in the Or- Data for Libraries, Archives and Museums: How to Clean, ganization of Knowledge, ed. Carol A. Bean and Rebecca Link and Publish Your Metadata. Chicago: Neal-Shuman. Green. Dordrecht: Springer, 3-18. Zeng, Marcia Lei. 2019. “Semantic Enrichment for En- Harpring, Patricia. 2018. “Linking the Getty Vocabularies: hancing LAM Data and Supporting Digital Humanities. The Content Perspective, Including an Update on Review article.” El profesional de la información, 28, no. 1. CONA.” In Human Rights in Cyperspace: Proceedings of the doi:10.3145/epi.2019.ene.03 2018 Pacific Neighborhood Consortium Annual Conference and
Knowl. Org. 46(2019)No.8 615 V. Clavier. Knowledge Organization, Data and Algorithns: The New Era of Visual Representation
Knowledge Organization, Data and Algorithms: The New Era of Visual Representations* Viviane Clavier GRESEC—Université Stendhal Grenoble 3—Institut de la Communication et des Médias, 11, avenue du 8 mai 1945—BP 337—38434 Échirolles Cedex,
Viviane Clavier is Associated Professor and Research Director in information and communication sciences. She is a member of Gresec at the University of Grenoble Alpes and her research focuses on knowledge organization and information seeking in the field of specialized and professional information. These studies focus on the themes of health and nutrition.
Clavier, Viviane. 2019. “Knowledge Organization, Data and Algorithns: The New Era of Visual Representation.” Knowledge Organization 46(8): 615-621. 35 references. DOI:10.5771/0943-7444-2019-8-615.
Abstract: This article shows how visual representations have progressively taken the lead over classical language- based models of knowledge organization (KO). The paper adopts a theoretical and historical perspective and focuses on the consequences of the changes in the volume of data generated by data production on the KO models. Until now, data visualization tools have been used mainly by researchers with expertise in textual data processing or in computational linguistics. But now, these tools are accessible to a greater number of users. Thus, there are new issues at stake for KO, other professions and institutions for gathering data that contribute to defining new standards and KO representations.
Received: 30 September 2019; Revised: 4 November 2019; Accepted: 4 November 2019
Keywords: knowledge, data, visual representations, images, information
* Presented at ISKO-France 2019: Open Data and Megadata in Social Sciences: New Challenges for the State and Organization of Knowledge?, at Université Paul-Valéry, Montpellier 3, France, October 9-11, 2019.
1.0 Introduction the reasons why languages are being called into question, and—if they are being replaced—by what and how. This This contribution is concerned with the place of languages study has a diachronic rather than historical dimension, in the process of collecting, representing and organizing which can be seen via research work in information and knowledge in a context characterised by the massive data communication sciences (CPDirSIC 2018), and it also production. By suggesting in the title that visual represen- makes reference to our own research in which language, tations are becoming crucial in representing knowledge, texts and discourse have always played a central role (Cla- we implicitly mean that they have supplanted languages. vier 2014). Our presentation is in two parts: the first part We are not saying that languages have disappeared, as the shows how visual representations have progressively taken use of controlled languages is still relevant to heritage in- over from language-based representations in knowledge stitutions and several fields of specialized information. organization systems. The second part shows how this However, the attention paid to languages has shifted to trend observed in research has also spread to civil society other knowledge objects (algorithms), other semiotics and to professionals concerned by information and com- (data visualization), other levels of representation (textual munication. However, it is clear that languages are still very data), other purposes of knowledge organization (brows- present, even if they appear to be relegated to a back- ing rather than categorization) and other properties (trans- ground descriptive function. parency and simplicity rather than precision and exhaust- iveness). It is important to notice that the term “language” is not mentioned in the last report published by the IFLA (2018), which is revealing about new trends in information concerning libraries. This paper, therefore, raises a series of questions, puts forward certain hypotheses, looks into 616 Knowl. Org. 46(2019)No.8 V. Clavier. Knowledge Organization, Data and Algorithns: The New Era of Visual Representation
2.0 The challenges of knowledge organization: were separated by a wall” (Van Slype 1982, 87). Viewed development and permanence from the logico-syntactic perspective of knowledge that could be processed by computer programs, natural lan- 2.1 Applied perspective: languages for knowledge guages were the subject of theoretical debates and consid- organization systems erable socioeconomic stakes. The questions raised were as follows (Lallich-Boidin et Maret 2005): 1) What is the rel- The notion of “languages” will be restricted to symbolic evant unit of meaning: the word or the morpheme?; 2) systems chosen to encode units of language and meanings, What advantages and disadvantages do free languages have which are used to feed knowledge organization systems in comparison with controlled languages?; and, 3)How can (KOS), “tools intended to organize recorded knowledge re- meaning be represented accurately while taking into ac- sources” (Sosińska-Kalata 2012). KOSs include “all kinds of count ambiguities and implications? It is necessary to use organization schemes ranging from simple alphabetical or external resources in order to define indexing languages slightly structured lists (authority records, glossaries, diction- (thesauri, controlled vocabularies) or to the documents aries, parts lists, etc.) to hierarchical classification schemes themselves—a question that raised the issue of faithful in- (classification plans, general or specialised classifications, dexing in light of cost savings in order to maintain lexical taxonomies, subject headings lists, etc.) or organizations em- resources. This approach conceived languages as being phasising relations that are not exclusively hierarchical (the- worked with and by computational linguists, and ques- sauruses, semantic networks, ontologies, etc.)” (Polity et al. tioned in relation to language units. It focused on a detailed 2005, 13). This perspective places us immediately in what description of language, based on a compositional ap- should be referred to as “the narrow sense” of knowledge proach to meaning led by syntax in the form of logical organization, namely the creation of information products representations, and was replaced by another approach intended to classify, organize and structure knowledge in or- based on an empirical representation of meaning in which der to facilitate access to information. This perspective is the aim was no longer to understand the language but to situated between a “broad” and theoretical conception of describe textual corpora. knowledge—for an ontological (Gnoli 2008) or social The 1990s thus represented a turning point character- (Hjørland 2015) conception of knowledge—and an “inter- ised by the web development, paving the way to corpus mediate” approach centered on the applications and forms linguistics (Habert et al. 1997), an approach that shifts the of knowledge mediation that can be observed in profes- focus towards the creation and description of textual data. sional situations (Clavier and Paganelli 2013, 2015). And so began the period of “large corpora” intended for What is common to documentary languages, classifica- annotation by exploration tools as a means of assessing tions, folksonomies, ontologies, etc. and today’s “data” is the representativeness of language phenomena. From this that they use processing methods in order to represent con- standpoint, emphasis was again put on the “vast” character tent. Thus, at any point in time, knowledge is the result of of data and text annotation or “tagging” aimed to collect choices and transformations governing its selection, per- typologies based on correlations of linguistic features haps its standardization and its categorization. The huge (Biber 1993). These annotation methods, which were often quantity of data to be processed is nothing new—even in manual or semi-automatic, combined with more robust in- the 1970s the development of document processing was formation processing methods such as learning-based text based on the idea that automation would help reduce the mining, aimed to perform supervised or unsupervised au- costs of manually indexing the constantly increasing docu- tomatic classification tasks. The knowledge resulting from mentation that professionals could no longer handle corpus annotation can be morpho-syntactic categories (Chaumier 1982). What we intend to question here is the (Poudat et al. 2006) or collocations (Tutin 1997), with nature of the knowledge selected with regard to computer phraseology having benefited from the joint development processing choices. of symbolic methods and probabilistic methods via auto- matic learning. Thus, unlike knowledge produced by com- 2.2 From the 1980s to today: the end of languages? putational linguistics, which aims to automate applications, the knowledge resulting from annotations produced by From 1980 to 1990, the development of full text infor- corpus linguistics “is based on an iteration between the mation systems and other applications such as linguistic analysis of computer outputs and ‘human’ consultation of technology-based knowledge extraction (Chaudiron 2007), texts or fragments of texts, for example concordances” enabled the analysis of languages from the standpoint of (Valette and Egle 2014). The production of knowledge linguistics (Van Slype 1982). The encounter between com- thus forms part of a tool-based approach aimed at explor- putational linguistics and documentation (Rouault 1987) ing data in order to pinpoint linguistic recurrences (idioms, brought together “these two disciplines even though they expressions, etc.). This perspective led to the development Knowl. Org. 46(2019)No.8 617 V. Clavier. Knowledge Organization, Data and Algorithns: The New Era of Visual Representation of methods based on “opportunistic” linguistic ap- erating indicators, maps, etc.” (Reymond 2016, 11). In the proaches, i.e., those that exploit automatically identifiable latter, the produced visual representations form the mate- surface phenomena. rial for computer graphics, laid out in such a way as to form Over time, the empirical approach became entrenched a language in its own right. However, what is visualized and enthusiasm for numerical processing methods in- may be data, information or knowledge. “Data visualiza- creased. The development of data analysis tools and the tion” is the generic term used to designate the tools of generalisation of methods intended for social sciences and “dataviz,” involving “a semiotic transformation between the “general public” was accompanied, as Dominique the results of a data analysis (numerical, categorial, textual) Boullier points out (Boullier 2015), by “the popular belief and a graphic representation” (Hachour 2015). The “visual that announced the end of science and scientific theories.” representation of information” is a branch of computing Thus began a new period in which big data “would be the that “uses visual and interactive representations of abstract effective measurements of reality” (ibid.). According to this data on a computer to amplify cognition” (Lamy 2017, 76 view, data are based on the collection of bags of words ff) and has six possible goals: displaying a large quantity of obtained from the web that are characterised by their vol- data, facilitating the search for given information, detect- ume and can be visually represented in two-dimensional ing patterns, enabling visual inferences, monitoring the oc- spaces. It is then a matter of interpreting and discovering currence of events and exploring data sets. The graphic this knowledge, the status of which is not always very clear; representations obtained in this way serve several applica- does knowledge represent themes, subjects, key words, dis- tions intended to visualize information, such as journalistic courses, lexicometric universes, etc.? The drawback of or documentary information. Thus, data visualization for these tools often lies in the opacity surrounding the data information (Yikun et Zhao 2016) is a means of discover- pre-processing stages, the choice of categories (both lexi- ing journalistic news by establishing “new interactions” cal and textual) or units of discourse. As pointed out by and presenting them visually. As far as documentary infor- Pascale Sébillot in 2002, mixed methods have indeed been mation is concerned, it may be visualized by maps that help developed, but in limited fields where the aim is no longer in browsing classifications (Dewey), directories (Rameau) to understand texts in detail but to obtain representations or catalogues (Papy 2005). Lastly the “visual representation of meanings that are useful for precise applications (Sebil- of knowledge” falls within the field of research into artifi- lot 2002). Thus, rule-based systems are used to develop re- cial intelligence and is based on the representation, model- sources for analysing sentiments and opinions (Poibeau ling and visualisation. According to the computer science 2014), so that it is possible to annotate data intended to researcher Jean-Baptiste Lamy (Lamy 2017, 12), a distinc- train classifiers. Other applications also rely on the profu- tion must be made between the iconic visual representa- sion of data, such as machine translation, data journalism tion of knowledge, which involves translating knowledge and fact checking. However, as pointed out by Thierry by using an iconic language including a pictogram glossary Poibeau (ibid.), the part connected with computational lin- and a grammar, and a structural visual representation of guistics is variable, and “the recurrent difficulties concern knowledge, which involves representing the structure of system adaptability, the time required to develop resources knowledge graphically using the techniques of visual rep- for a new field and the availability or lack of a sufficient resentation. number of examples.” The researcher also indicated that other frequent questions concern the definition of the in- 3.0 Knowledge dissemination: priority of visual formation sought and the quality of the results obtained. representations for many actors For a number of years now there has been considerable interest in visual representation, as revealed by a recent re- The choice of verbal or visual semiotics to represent view of works on the subject carried out by researchers at knowledge in information systems is not simply a matter the University of Swansea (Rees and Laramee 2019). The for science but falls within a much wider spatio-temporel, authors’ research goes back as far as 1967, to the Semiology socio-economic and cultural context. Visual representa- of Graphics by Jacques Bertin, one of the first monographs tions are now the predominant modes of expression in our devoted to the question. According to them, the following society, as evidenced by computer graphics, fixed and ani- forty years were not particularly prolific, but since the early mated images (videos) in the field of public and scientific 2000s there has been a plethora of publications, to the communication, in the media and on the web. To remove point where it is difficult to keep track of them (610). Vis- any ambiguity, it is not a question of demonstrating that ual representation must be understood as both a process we live in “a world of images” as suggested by the philos- of creating visibility and as the result of that process. In opher Franck Robert in his introduction to the joint pub- the former case, it is “the result of the instrumented ex- lication “Philosophies de l’image,” but rather that the im- ploration of masses of data that become suitable for gen- portance assumed by visual representations—or “visuals” 618 Knowl. Org. 46(2019)No.8 V. Clavier. Knowledge Organization, Data and Algorithns: The New Era of Visual Representation as they are referred to by communication specialists—as little pebbles. If they are scattered intelligently, they forms of semiotization influences the ways in which will attract far more traffic to a website than a simple knowledge is represented and organized. The rest of this ‘quote.’ article will present a few trends observed among actors who, for one reason or another, play an important role in When you realise that photos posted on Facebook can the dissemination of knowledge representation standards generate 53% more ‘likes’ on a post, you make the or formats. These examples—which are not derived from most of them. Users are far more inclined to like or a reasoned corpus—illustrate the fact that knowledge is al- follow a brand if it is active and shares images. Textual ways rooted in particular periods and places and probably promotion is a thing of the past. depends on fashions. Web writing standards for the social media now advise 3.1 Promotion of images in web publication tools choosing visual semiotics: it is better to “show” than to and techniques “tell,” words are “there to underscore the image,” “visual stratagems are to be used to win visitors,” etc. Even so, the Images are promoted by the web professionals who define eternal question of document description, representation the writing standards and techniques aimed at optimising and indexing is still raised in the context of information the visibility of websites via their referencing and position- searches in the form of user requests. Ultimately, then, im- ing in search engines. There are very many sites aimed at ages always lead back to the question of languages and the web publishers giving recommendations on how to write choice of language to describe them. This question is dealt for the web and setting out rules on concision, simplicity, with more specifically on sites intended to enhance the vis- content structure, addresses, etc.1 With regard to images, ibility of websites in search engines (search engine optimi- they may be fixed or animated, such as videos. On the web- zation) or on the social media (social media optimization) site Annei, images and text, “always considered to be ri- or specifically in engines dedicated to marketing (search vals,” are now presented as complementary, subject to a engine marketing). Reference is made to W3C, which de- few adjustments intended to prevent images being re- fines web languages and standards and gives the image for- stricted to a secondary role. The site advises combining mats accepted by search engines, description tags and ad- “alternative text,” a “concise, descriptive textual equiva- vice on describing the alt and title attributes for presenting lent” along with images that do not exceed “250 charac- “the content of an image clearly and concisely.” ters,” or “ensuring that video soundtracks will be made ac- cessible by a retranscription in text format that can be de- The alternative text of images: Google also references tected by search engines.” As far as the choice of images images! Always remember to complete the alternative is concerned, it also gives recommendations on their con- text (alt tag) of the images in your articles. Introduce tent: “Choose images that have an informative value.” This your main keyword within a short expression. The alt advice is aimed at eliminating images that are purely illus- tag regularly omitted is that of the image on the first trative. Images considered to be informative include: page, which weakens image referencing. To avoid this, “computer graphics, diagrams, photographs taken in real fill in the Alternative text field when you define your situations, providing more than a visual taken from a data first page image. bank and is of purely decorative value” (ibid.). In this last instance, it advises adding “a concise legend that gives It is clear that images, far from replacing text, once again meaning to the image, whenever possible,” in order to re- raise the question of textual representation, the choice of inforce “the image’s impact rating.” words and their descriptive virtue, reversing the roles of “il- The advice given with regard to publishing on the social lustrative” images in favour of text. From the point of view media is that images should even replace text on the of information searching, these trends have led to the de- grounds that an image simply requires a click to score a hit, velopment of research into the automatic classification of thus guaranteeing traffic that generates profits or viewer- images, facial recognition, and the production of languages ship. This recommendation appears to be connected with for describing and annotating images, such as ontologies. In- users’ observed habits: they click on images to browse the deed the “Google Image” search engine is becoming in- internet: creasingly popular with users to the extent that natural image referencing now appears to be a priority.2 Pinterest and Facebook users are accustomed to zig- zag browsing. One click leads to another click, an im- age leads to a link where there are other images, which also lead to other links. Images are like Tom Thumb’s Knowl. Org. 46(2019)No.8 619 V. Clavier. Knowledge Organization, Data and Algorithns: The New Era of Visual Representation
3.2 Development of data visualization tools for the success would appear to be due to their transparency, sim- general public plicity and faithfulness to the data, as can be seen in this example of an announcement presenting the advantages Hitherto reserved for specialists in multidimensional de- of dataviz for representing textual or numerical data. scriptive statistics, certain methods such as principal com- ponent analysis (PCA) and correspondence factor analysis Dataviz, or data visualization, is a practice that we (CFA) have been widely used in the humanities and social encounter on a daily basis without realising it, just by sciences to study literary or political texts from different opening a newspaper or watching the television. The viewpoints: lexicon, vocabulary, style, themes, etc. An ex- simplest example is the survey. With the digital age, tensive literature devoted to textual statistics has been built it has become a powerful communication tool. up since the late 1950s, the common theme of which is the production of knowledge organization systems in the Dataviz: a definition that is easy to understand. form of verbal semiotics: dictionaries (Guiraud 1959), in- dexes, concordances, repeated segments, textual time se- In a society that is increasingly attracted to graphics, ries, parallel corpora, etc. (Salem et Lebart 1996). In these data visualization takes precedence over raw data. It works, visual representations—entire lexical tables, figures helps to throw light on information that is appar- representing factorial planes, etc.—are not a goal in them- ently complex or is submerged in a large quantity of selves but are considered to be an “aid in reading a series parameters. The term dataviz thus refers to a set of of texts” (135 ff). These introductory works give guide- visual representations of these raw data. What is its lines on: “How to interpret distances,” “Examples of ap- principal objective? To throw light on a phenome- plications,” “Reading a figure,” etc. It is thus clear that the non by giving it a more ergonomic appearance than interpretation of textual data is subject first of all to lexi- a spreadsheet filled with figures. cometric reorganization in the form of visual representa- tions, which are the result of a set of processing operations Transparency refers to the immediately perceptible and un- and choices concerning the size of the textual corpora, derstandable character of a message when it is represented their representativeness and segmentation, the definition in a visual form, which would give it undeniable communi- of counting units, etc. As a follow-up to these methods, cational virtues. Simplicity is the ability of the tools to pro- text mining includes statistical data processing and tasks vide a summary representation in visual form of infor- such as supervised or unsupervised classification (Ibekwe- mation that is expressed in a verbal manner, which is thus SanJuan 2007). Developed by computer scientists special- felt to be more complex. Lastly, faithfulness is the ability of ising in learning, these methods approach texts as data in- the tools to represent data such as that emanating from in- tended to train classifiers, whose scores are measured and dividuals without any form of mediation. These properties, compared. Applied research as information retrieval, pat- which are presented as technical assets, are the subject of tern detection, scientific monitoring, etc. are always envis- ongoing theoretical debates on the differences in semiotic aged, so that the text in the strict sense and considered in status of pictorial and written signs (Christin 2012). These its production conditions, is not the focus of attention. It arguments hark back to the sterile debates on the more is this transition from viewing text as sacred to viewing it transparent character of pictorial signs in comparison with as an instrument, a training corpus, that appears to be at written signs on the grounds that there is no need to use an the origin of an interpretation centred on understanding arbitrary code to interpret an image, in contrast to a text; trends, regularities, mass effects, etc. and less on what is of likewise simplicity no doubt refers to the small number of the order of the invisible, which can only be retrieved signs used, as a small set of visual forms is always more con- through knowledge of the texts. cise than an infinite number of verbal signs combined to At the same time, dataviz tools are becoming far more describe a language. These debates are far from over, and commonplace as many actors require analyses of discourse they question the pertinence of knowledge encoding pro- on the social media, blogs, Twitter, forums, etc. Thus, for cesses without taking into account reception conditions. economic reasons, survey institutions have introduced so- Lastly, faithfulness is also a question that is widely discussed called passive methods, inspired from big data, which are in the information-documentation field with regard to the qualitative, can be automated and used to complement merits of free and controlled indexing languages. quantitative suveys. There is a wide social demand: market- The representation of knowledge in visual form is also ing, political polls, consumption, etc. There is a multitude of interest to communication specialists who approach in- of websites that list dataviz tools, distinguishing between formation design from the aesthetic standpoint. They are various families: mental maps, visual representation of re- required to posses new dataviz skills at the crossroads be- lational data networks, information mapping, etc. Their tween computing, statistics and communication (Arruabar- 620 Knowl. Org. 46(2019)No.8 V. Clavier. Knowledge Organization, Data and Algorithns: The New Era of Visual Representation rena 2017). Lastly, the visual representation of knowledge Supported by specialist disciplines in big data processing has been lifted to the status of an artistic object worthy of and encouraged by research and education policies, data vis- being displayed. For example, the temporary exhibition at ualization is more than just a tool for the knowledge man- the Mundaneum in Mons entitled “Mapping Knowledge” agement profession, it is a genuine aid to discovering presents the work of Paul Otlet, who is more famous for knowledge. It is no longer a matter of naming knowledge in having created the universal decimal classification system order to understand it, but rather of showing it in order to than for his texts and drawings aimed at “summarising what make it visible, a perspective that leaves considerable leeway is known” (see Figure 1). for new actors, new specialisations and new data collection structures that are helping to redefine standards, forms of 4.0 Conclusion writing and methods for representing knowledge.
Focusing on the place of visual representations as know- Notes ledge organization systems is part of a long line of research into the choice of vocabularies and how they are organized 1. The examples of sites mentioned here are among the and structured in order to access knowledge. This contribu- top results obtained by Googling “web image and text tion illustrates the fact that visual representations obtained publication.” through data processing are neither “natural,” nor “simple,” 2. https://www.1ere-position.fr/blog/optimiser-le-refer nor are they “objective” even if they are made by automatic encement-dimage-sur-google-images/#Google_Im tools and are applied to already existing data (Bullich and ages_en_forte_croissance (consulted on 2 May 2019). Clavier 2018). This observation was made in particular by Michèle Hudon and Widad Mustapha El Hadi with respect References to classifications: “There is no natural organization structure and any classification or structure can only be invented and Arruabarrena, Béa. 2017. “L’expert en dataviz, un métier built, never discovered. Furthermore, the establishment of en transition.” I2D – Information, données & documents 54, a classification framework or structure is always evidence of no. 3: 7-8. an encounter or clash between philosophical, technological, Biber, Douglas. 1993. “Using register-diversified corpora for social, economic and political imperatives (Svenonius general language studies.” Computational Linguistics 19: 1991)” (Hudon and Mustafa El Hadi 2010, 12). 243-58.
Figure 1. Screen capture of the exhibition devoted to the visual representations of Paul Otlet, obtained from GoogleArts& Culture. Knowl. Org. 46(2019)No.8 621 V. Clavier. Knowledge Organization, Data and Algorithns: The New Era of Visual Representation
Boullier, Dominique. 2015. “Les sciences sociales face aux IFLA. 2018. Riding the waves or caught in the tide? Navigating the traces du big data. Société, opinion ou vibrations ?” Re- evolving information environment. Insights from the trend report. vue française de science politique 65: 805-28. Lallich-Boidin, Geneviève and Maret, Dominique. 2005. Re- Bullich, Vincent et Clavier, Viviane. 2018. Dossier 2018. cherche d’information et traitement de la langue: fondements linguis- “Production des données, ‘production de la société’. Les tiques et applications, Villeurbanne: Presses de l’Enssib. Big Data et algorithmes au regard des sciences de l’infor- Lamy, Jean-Baptiste. 2017. Représentation, iconisation et visua- mation et la communication.” Les Enjeux de l’information et lisation des connaissances: Principes et applications à l’aide à la la communication. décision médicale. Habilitation à diriger des recherches en Chaudiron, Stéphane. 2007. “Technologies linguistiques et informatique, Université de Rouen. modes de représentation de l’information textuelle.” Do- Lebart, Ludovic et Salem, André.1994. Statistique textuelle. cumentaliste-Sciences de l’Information 44: 30-39. Paris: Dunod. Chaumier, Jacques. 1982. Analyse et langages documentaires. Le Papy, Fabrice. 2005. “Au-delà de la transfiguration du cata- traitement linguistique de l’information documentaire. Paris: En- logue.” Bulletin des bibliothèques de France 4: 5-12. treprise moderne d’édition. Poibeau, Thierry. 2014. “Le traitement automatique des Christin, Anne-Marie. 2012. Histoire de l’écriture. De l’idéo- langues pour les sciences sociales. Quelques éléments gramme au multimedia. Paris: Flammarion. de réflexion à partir d’expériences récentes.” Réseaux Clavier, Viviane. 2014. L’organisation des connaissances au prisme 188, no. 6: 25-51. du langage, du texte et du discours. Un parcours en recherche d’in- Polity, Yolla, Henneron, Gérard et Palermiti, Rosalba. (ed.) formation. Habilitation à diriger des recherches en sciences 2003. L’organisation des connaissances. Approches conceptuelles. de l’information et la communication, Université Sten- Actes du 4e congrès ISKO-France, Paris: L’Harmattan. dhal-Grenoble3. Poudat, Céline, Cleuziou, Guillaume et Clavier, Viviane. Clavier, Viviane et Paganelli, Céline. 2015. “Activités infor- 2006. “Catégorisation de textes en domaines et genres: mationnelles et organisation des connaissances: résultats complémentarité des indexations lexicale et morpho- et perspectives pour l’information spécialisée.” Les cahiers syntaxique.” Document numérique 9: 61-76. de la Société Française des Sciences de la Communication, 11: 170- Rees, Dylan and Laramee, Roberts S. 2019. “A survey of 75. Information Visualization Books.” Computer Graphics fo- Clavier, Viviane et Paganelli, Céline. 2013. L’information pro- rum 38: 610-46. fessionnelle. Paris: Lavoisier Hermès Sciences Publications. Reymond, David. 2016. “Introduction. Visualisation de CPDirSIC (Conférence Permanente des Directeurs·trices données.” Les Cahiers du numérique 12, 4: 9-18. des unités de recherche en Sciences de l’Information et Rouault, Jacques. 1987. La linguistique automatique. Applications de la Communication). 2018. Dynamique des recherches en documentaires. Berne: Peter Lang. sciences de l’information et de la communication. Sébillot, Pascale. 2002. Apprentissage sur corpus de relations lexi- Gnoli, Claudio. 2008. “Ten Long-Term Research Questions cales sémantiques. La linguistique et l’apprentissage au service d’ap- in Knowledge Organization.” Knowledge Organization 35: plications du traitement automatique des langues. Habilitation à 137-49. diriger des recherches en informatique, Université de Guiraud, Pierre. 1959. Problèmes et méthodes de la statistique Rennes 1. linguistique. Dordrecht: D. Reidel. Sosińska-Kalata, Barbara. 2012. “L’efficacité des systèmes Habert, Benoit, Nazarenko, Adeline et Salem, André. d’organisation des connaissances: un point de vue 1997. Les linguistiques de corpus. Paris: Armand Colin. praxiologique.” Études de communication 39: 155-71. Hachour, Hakim. 2015. “De la fouille à la visualisation de Tutin, Agnès. 1997. “Coder les collocations dans un lexique données: un processus interprétatif.” I2D – Information, formel pour le TALN.” Revue française de linguistique appli- données & documents 52, no. 2: 42-43. quée, vol. ii: 43-58. Hjørland, Birger. 2015. “Theories are knowledge organizing Valette, Mathieu et Egle, Eensoo. 2014. “Approche textuelle systems (KOS).” Knowledge Organization 42: 113-28. pour le traitement automatique du discours évaluatif.” Hudon, Michèle et Mustafa El Hadi, Widad. 2010. “Organi- Langue française 184: 109-24. sation des connaissances et des ressources documen- Van Slype, Georges. 1977. Conception et gestion des systèmes do- taires. De l’organisation hiérarchique centralisée à l’orga- cumentaires. Paris: Les éditions d’organisation. nisation sociale distribuée.” Les Cahiers du numérique 6, no. Yikun, Liu et Zhao, Dong. 2016. La datavisualisation au service 3: 9-38. de l’information. Paris: Pyramid. Ibekwe-Sanjuan, Fidelia. 2007. Fouille de textes. Méthodes outils et applications, Paris: Lavoisier Hermès Sciences Publica- tions.
622 Knowl. Org. 46(2019)No.8 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
Data Papers as a New Form of Knowledge Organization in the Field of Research Data† Joachim Schöpfel*, Dominic Farace**, Hélène Prost***, and Antonella Zane**** *University of Lille, GERiiCO laboratory, ORCID 0000-0002-4000-807X,
Joachim Schöpfel is Associate Professor in information and communication sciences at the University of Lille (France). He is a scientist at the GERiiCO laboratory and an independent consultant with the consulting com- pany Ourouk, Paris. He has been teaching at different universities (Sorbonne, Lyon, Köln, Ljubljana), he is pro- fessional trainer of academic librarians and documentalists and of DARIAH-EU masterclasses Digital Human- ities (Prague 2016, Paris 2017), and he is member of several editorial boards and conference committees.
Dominic Farace heads GreyNet International and is Director of TextRelease, an independent information bu- reau specializing in grey literature and networked information. His doctoral dissertation in social sciences is from the University of Utrecht, where he has lived and worked since 1976. Farace founded GreyNet, Grey Literature Network Service in 1992. He has since been responsible for the International Conference Series on Grey Liter- ature and serves as Program and Conference Director as well as managing editor of the Conference Proceedings. He is the editor of The Grey Journal and provides workshops and training in the field of grey literature.
Since 1995, Hélène Prost has been an information professional at the Institute of Scientific and Technical Infor- mation (CNRS). After participating in the management of printed subscriptions, she conducted studies on the development of databases. Currently she is conducting statistical studies on the use of electronic resources of the CNRS portal. Since 2011, she has been an associate member of the GERiiCO research laboratory (University of Lille). She is responsible for the development of the collection GERiiCO on HAL. She participates in research projects on evaluation of collections, document delivery, usage analysis, grey literature, and open access.
Antonella Zane has a PhD in earth sciences and ten years of research background at the University of Padova. Since 1998, she has been working for the Padova University Library System, and from 2002 she has co-operated with the Digital Library Group of ISTI-CNR in Pisa. Antonella Zane was Project manager of the EU Project “Linked Heritage.” From 2011 to 2018 she has been responsible for the Digital Library of Padova University with special focus on research data repositories. At present, Antonella Zane is Head of Vallisneri Bio-Medical Library of University of Padova and a member of the Italian Open Science Support Group.
Schöpfel, Joachim, Dominic Farace, Hélène Prost and Antonella Zane. 2019. “Data Papers as a New Form of Knowledge Organization in the Field of Research Data.” Knowledge Organization 46(8): 622-638. 27 references. DOI:10.5771/0943-7444-2019-8-622.
Abstract: Data papers have been defined as scholarly journal publications whose primary purpose is to describe research data. Our survey provides more insights about the environment of data papers, i.e., disciplines, publish- ers and business models, and about their structure, length, formats, metadata, and licensing. Data papers are a product of the emerging ecosystem of data-driven open science. They contribute to the FAIR principles for research data management. However, the boundaries with other categories of academic publishing are partly blurred. Data papers are (can be) generated automatically and are potentially machine-readable. Data papers are essentially information, i.e., description of data, but also partly contribute to the generation of knowledge and data on its own. Part of the new ecosystem of open and data-driven science, data papers and data journals are an interesting and relevant object for the assessment and understanding of the transition of the former system of academic publishing.
Received: Revised: Accepted: 3 December 2019
Keywords: data, papers, journals, research, metadata
† Presented at ISKO-France 2019: Open Data and Megadata in Social Sciences: New Challenges for the State and Organization of Knowledge?, at Université Paul-Valéry, Montpellier 3, France, October 9-11, 2019. The paper builds on the scientific and professional Knowl. Org. 46(2019)No.8 623 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
experience of all authors in the framework of the international Grey Literature Network Service1 and its international workshops on data papers (Netherlands, Czech Republic, Italy, and the United States), its collection of data papers in the Dutch EASY data repository, and its follow-up study of enhanced publications.
1.0 Introduction and discussion of these results (papers), and the descrip- tion (cataloguing, abstracting, and indexing) of those da- In the context of open science, an increasing volume of tasets and papers. This emerging category of data papers research data are made available on the internet, contrib- appears to challenge this clear distinction, interlinking da- uting to the so-called big data of science. New tools, meth- tasets, papers and metadata, blurring boundaries, changing ods, and infrastructures have been developed for the dis- priorities, and modifying the basic purpose of academic semination, processing, analysis, and preservation of re- publishing. search data. Data papers are part of them. Built on an overview of recently published studies, the Data papers are a young species of academic publishing. following study produces an empirical update on the pub- In 2006, Pärtel stated (99) for the field of ecology that “until lishing of data papers: the number and development of now ... very few data papers have appeared.” In fact, most data papers and journals, the country and language of pub- of the data papers or papers about data papers have been lications, the platforms and publishers, as well as the busi- published since 2008 and 2009.2 Yet, as Smith (2012, 16) re- ness models. The purpose of our paper is to analyse data minds, “the concept has actually been around for quite a papers as a new tool of scientific communication and to while (even if) the older journals that date from the print era produce insight on their contribution to the organization tend to be not particularly useful in the modern environ- of scientific knowledge via questions pertaining to the ment.” In fact, one (the first?) data journal (Journal of Chem- production and the functions of data papers: ical and Engineering Data From ACS) was already launched in 1956 (see the timeline in Garcia-Garcia et al. 2015). – How are they “written”? The simplest definition (Callaghan et al. 2012, 112) is that – Which is the link with data repositories, metadata, and data papers focus on “information on the what, where, why, other papers? how and who of the data” rather than original research re- – Which is the (potential and real) part of automatic or sults. semi-automatic production? Data papers have been defined as “a searchable metadata – Which is the part of human added value? document, describing a particular dataset or a group of da- – Which degree of standardization, which link between tasets, published in the form of a peer-reviewed article in a metadata formats and the data journals’ author guide- scholarly journal.”3 They are published in specific data jour- lines? nals like Data in Brief (Elsevier) and Scientific Data (Nature) or – In which way are data papers related to the so-called in regular academic journals with special sections for data “FAIR Guiding Principles for Scientific Data Manage- papers, like BMC Research Notes GigaScience (Oxford Univer- ment and Stewardship” (Wilkinson et al. 2016)? sity Press) and PLoS One. Most data papers are published in – Do they just improve the referencing of datasets on re- journal platforms; yet, some are (also or exclusively) pub- positories, or do they fulfil other roles? lished on data repository platforms.4 Unlike usual research – Are data papers “written by machines” and meant to be papers, the main purpose of data papers is to describe da- “read by machines?” tasets, including the conditions and context of their acquisi- tion and their potential utility, rather than to report and dis- The paper will conclude with a conceptual approach to cuss results. Also, it is generally assumed that data papers are data papers as part of the organization of knowledge short papers with up to four pages. based on research data in the context of open science. In the “classical” research paradigm, the focus is on ar- ticles presenting results while research data are useful for 2.0 Literature overview the validation of published research findings. Data papers invert the roles, insofar the paper’s main function is to in- 2.1 Definitions and functions form about and link to research data on data repositories, contributing to their findability and reusability. Are data An increasing number of journal editors announce the papers complementary to research papers, or will they re- launch of a new section with data papers. They put for- place them, as a seamless and direct way of providing ac- ward different objectives, even if the main purpose is sim- cess to research results? ilar: to inform about research data and to foster their ac- Also, traditional knowledge organization makes a clear cessibility and reuse. Three examples among others illus- distinction between research results (datasets), the analysis trate the diversity of goals: 624 Knowl. Org. 46(2019)No.8 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
– The objective of The International Journal of Robotics Re- ricultural and biological sciences (16%). They identified search is “to facilitate and encourage the release of high- only nine data journals in social sciences and humanities quality, peer-reviewed datasets to the … community” (8%). Their results show a recent and slowly developing (Peter and Corke 2009, 587). landscape (the average number of data papers per journal – Studies in Family Planning tries to promote “interdisci- is less than ten), with conceptual, structural and termino- plinary research and integrative analyses by making ac- logical diversity (they identified ten different terms as- cessible to researchers, policymakers, students, and do- signed for data papers). Also, there is no consensus about nors data that may be useful in answering critical ques- the usual content; the only section present in all data pa- tions of interest to … readers” (Friedmann et al. 2017, pers is the data availability (location, accessibility), followed 291). by information about the provenance of the dataset. Most – The French journal of information and communication of the data journals perform some kind of traditional peer sciences RFSIC invites data papers to describe the sci- review to guarantee a certain level of the papers’ quality entific process, methods, and tools that result in re- but also to assert some quality of the datasets in terms of search data in a Bruno Latour perspective, “since they utility and reusability; only a few journals adopted an never just magically appear” (Le Deuff 2018, 2). “open peer review.” Most journals are published in open access, with an average APC5 amount of 1,300 euros. The publisher Pensoft describes (Penev et al. 2012) a data The Grey Journal, published by Textrelease (Amsterdam) paper as “a scholarly journal publication whose primary is one of those “mixed” data journals. Initially a regular purpose is to describe a dataset or a group of datasets, ra- journal with papers from international conferences and ther than to report a research investigation. As such, it con- original research articles, The Grey Journal started to publish tains facts about data, not hypotheses and arguments in a collection of data papers in 2017. This collection was support of the data, as found in a conventional research born out of an “Enhanced Publications Project,” fueled article.” by the FAIR Data Principles (Farace et al. 2018). The main The term remains ambiguous. For instance, Bordelon pillars for this collection are the International Conference et al. (2016, 1) define data papers as “papers that present, Series on Grey Literature, the research data that is created analyze, or use data obtained with the respective facilities” and archived within this framework (actually thirty-seven (i.e. observatories). Pärtel (2006) considers data papers as datasets housed in the Dutch data repository DANS), and a kind of “abstracts” that aim to collect, organize, synthe- the existence of a flagship journal for the publication of sise, and document data sets of value in a given field; only the data papers. A standardized template is provided to en- the abstract appears in a data journal (or the data paper sure the identity and longevity of the collection and to section of a regular journal) while the data and metadata guide prospective authors and researchers in submitting a are available through a field-specific data repository on the data paper. The template consists of five sections each of internet. For Penev et al. (2012), their purposes are three- which has a note field providing examples and/or a maxi- fold: “to provide a citable journal publication that brings mum word count. The fields are labelled as follows: over- scholarly credit to data publishers; to describe the data in view, methods, data description, potential reuse, and refer- a structured human-readable form; (and) to bring the ex- ences. Currently, seven of GreyNet’s thirty-seven datasets istence of the data to the attention of the scholarly com- are supported by a data paper (19%). Yet, even on a small munity.” At first sight, data papers, in spite of their com- scale this data paper collection illustrates an operational mon general purpose, appear to belong to a rather hetero- and functional ecosystem of open science constructed genous and dissimilar new kind of document. Our study year after year with five main elements, i.e., an academic will reveal, nevertheless, more common features, such as community, original research within this community, con- the fundamental structure. ferences, a journal, and a data repository. In this emerging framework, data papers gain their particular relevance, dif- 2.2 Data journals ferent from regular articles.
A first survey on data journals was conducted by Candela 2.3 Features and metadata et al. (2015), with a sample of 116 data journals published by fifteen different publishers. They distinguished seven Yet, other aspects appear challenging the idea of a clear “pure” data journals publishing only data papers and 109 distinction between data papers and regular papers. Li et “mixed” data journals publishing any typology of paper al. (2019) conducted a content analysis with eighty-two including data papers. The most represented subjects (in data papers from sixteen journals to investigate what in- terms of number of journals) were medicine (53%), bio- formation they describe regarding the methods to create chemistry, genomics and molecular biology (26%), and ag- and manipulate the data objects (i.e., “data events”). For Li Knowl. Org. 46(2019)No.8 625 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data and his colleagues, even if they have distinct features from ifiability, and credibility of those resources,” persistent research articles, data papers are “nevertheless created un- identifiers, an “interpretive analysis of the data (which) der similar conditions,” and they reveal “functional over- could include taxonomic, geospatial or temporal assess- laps” between both categories related to the narratives of ment of data and its potential of integration with other data events (natural language) and to their composition, types of data resources” or the inclusion of “a taxonomic which is “inevitably situated in the specific epistemic com- checklist and/or the data themselves.” Data papers repre- munities.” Their main function is to improve the findabil- sent a highly standardized type of publication, with a ity of published datasets and, through enriched metadata standard structure and a content, which is largely defined description, to foster their reusability. in terms of metadata formats (such as DataCite Metadata Metadata are constitutive for data papers. Candela et al. Schema) and identifiers for datasets, persons, etc. (such as (2015) produced a conceptual map of the data paper (see DOI and ORCID). Figure 1). They insist not to confuse the data paper’s con- tent, its metadata, and the datasets’ metadata. “The con- 2.4 Production and processing cept of data paper has at least two elements that have to be materialized into concrete and identifiable information In fact, Chavan and Penev (2011) describe an integrated objects in order to fully implement it: the dataset, i.e., the workflow of data repositories and journal platforms, re- subject of the data paper, and the data paper itself, i.e., the quiring shared standards and formats. Senderov et al. artefact produced to describe the dataset” (1752). (2016) provide an example of this data paper generation The link between data papers and the metadata of re- in the field of biodiversity. Their workflow relies on three search data is essential, because both have similar func- key standards (RESTful API’s for the web, Darwin Core, tions (describe data, define accessibility, (re)usability, and and EML) and imports metadata into the ARPHA writing content. Insofar data papers are about deposited datasets tool (AWT). In other words, and more generally spoken, and insofar deposits require metadata, data papers can be “the boundary between a workflow tool, a data store, and (partly) derived from existing metadata. a publishing platform blurs” (de Waard 2010, 9). Chavan and Penev (2011, 7) describe a tool that “facili- But are data papers produced only for machines? No, ac- tates conversion of a metadata document into a traditional cording to Li et al. (2019, 18) who are convinced that “as a manuscript for submission to a journal” for biodiversity genre built upon natural languages, data papers are primarily resource datasets. The human contribution is minimal if a human-readable document, much less designed for repro- the metadata is standardized (with controlled vocabulary), ducing data workflows in computational approaches.” Both exhaustive, and of sufficient quality: “Once the metadata are complimentary rather than competitive. are completed to the best of the author’s ability, a data pa- In her review of data papers, Reymonet (2017) compares per manuscript can be generated automatically from these data papers and data management plans (DMP). Indeed, as metadata using the automated tool … The author checks the expected structure of such an article may be based on the created manuscript and then submits it for publication the items provided when preparing a DMP, Reymonet sug- in the data paper section through the online submission gests a tool (or workflow) to export selected items of DMPs system of an appropriate … journal” (7). in order to prepare or generate a data paper. This kind of generated data paper can be further en- A general assumption is that data papers, like regular hanced in different ways, such as “describing fitness for papers, are peer reviewed, implying some kind of quality use of data resources, which will increase the usability, ver- control and selection. This means, too, that metadata of
Figure 1. Data papers concept map (from Candela et al. 2015, 1752). 626 Knowl. Org. 46(2019)No.8 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data research data (and, indirectly, the datasets themselves) be- 3.0 Methodology come the object of scientific evaluation, which “contrib- utes to the popularity of data papers in increasingly more In order to analyse specific features of data papers, we es- scientific fields” (Li et al. 2019, see also Costello et al. tablished a representative sample of data journals, based 2013). For the same reason, data papers contribute to the on lists from the European FOSTER Plus project,7 the trustworthiness of research data. For example, Elsevier’s German wiki forschungsdaten.org hosted by the Univer- Chemical Data Collections invites authors to submit data pa- sity of Konstanz,8 and two French public research organi- pers, because this “ensures that your data … is actively zations.9 The complete list consists of eighty-two data peer reviewed.”6 As cited above, Pärtel (2006) mentions journals, i.e., journals that publish data papers. They repre- that data papers were about “data sets of value in a given sent less than 0.5% of academic and scholarly journals. For field,” which implies a selection by the authors themselves, each of these eighty-two data journals, we gathered infor- upstream of the writing of data papers and of peer re- mation about the discipline, the global business model, the views even if the criteria of selection remain uncertain. publisher, peer reviewing, etc. The analysis is partly based on data from ProQuest’s Ulrichsweb database, enriched 2.5 Critics and outlook and completed by information available on the journals’ home pages. Similar to most cited authors, Smith (2012, 15) states that Some data journals are presented as “pure” data journals data papers “are like traditional research papers in some stricto sensu, i.e., journals which publish exclusively or mainly aspects: they are formally accepted, they are peer-reviewed, data papers. We identified twenty-eight journals of this cat- they are citable entities” but then adds that “in other re- egory (34%). For each journal, we assessed through direct spects they are very different from traditional research ar- search on the journals’ homepages (information about the ticles because they are not about the research, they are journal, author’s guidelines, etc.) the use of identifiers and about the data.” And this exactly is the main reason for metadata, the mode of selection and the business model, some more critical voices, expressing concerns about the and we assessed different parameters of the data papers real demand by society and research, about the additional themselves, such as length, structure, linking, etc. workload for authors and peer reviewers, and about the The results of this analysis are compared with other re- motivation of scientists to share their data. The underlying search journals (“mixed” data journals) that publish data idea is that scientists should (and mostly do) publish about papers along with regular research articles in order to iden- results not about data. tify possible differences between both journal categories Other arguments against data papers are their price on the level of data papers as well as on the level of the (APCs) and the slow uptake, at least initially. “To address regular research papers. Moreover, the results are discussed professional recognition and data quality control, there are against concepts of knowledge organization. viable alternatives to the data paper (such as the) imple- mentation of a joint data-publishing and -archiving policy 4.0 Results by databases and journals … instead of popularizing a new kind of publication, it is more important to improve cur- Four of the twenty-eight data journals have ceased, and rent peer-review processes and the operating policies and two have merged. All of them are published online while integration of journals and databases” (Huang et al. 2013, nine still have a print version. One data journal is a report 5). Huang’s critic may be specific for a given field of re- series. search (here, biodiversity) but should be taken into account for a general understanding of the future development of 4.1 Research disciplines data papers. Nevertheless, data journals and data papers appear to Most data journals are from STEM domains, in particular be here to stay. The French national plan for open science from life and medical sciences, including genetics (see Fig- recommends (MESRI 2018, 6-7) “as part of its govern- ure 2). Only four journals publish data from the humani- ment support for journals … the adoption of an open data ties (psychology, archaeology) and social sciences. One policy associated with articles and the development of data data journal covers a large range of disciplines from sci- articles and data journals.” While data papers become a le- ences (Scientific Data by Nature), another is open for all top- gitim (mainstream) part of the landscape of academic ics in social sciences and humanities (Research Data Journal publishing, only few studies provide empirical or concep- for the Humanities and Social Sciences by Brill). tual elements of an answer to the question of how exactly The five data journals with papers on data from the arts, data papers contribute to the organization of scientific social sciences, and humanities represent 18% of all knowledge, compared to regular research articles. “pure” data journals. In terms of articles (see below), they Knowl. Org. 46(2019)No.8 627 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
Figure 2. Number of data journals per domain (N=28). represent less than 4% of all data papers published in data 2000. The other twenty-one journals have been launched journals, with estimated 400-450 papers, mostly in archae- during the last ten years, from 2008 on, and especially in ology. 2013 (seven journals) and 2014 (five journals). Four jour- nals have ceased or are suspended. 4.2 Publishers At least one part of the data journals is considered as a good or high-quality journal. Eleven data journals are in- Except for Taylor & Francis, all big five academic publishers dexed by Clarivate Analytics and eight by Elsevier’s Scopus (Elsevier, Springer-Nature, Wiley-Blackwell, Taylor & Fran- database. Sixteen journals are referenced in the interna- cis, and SAGE) have their own data journals. Five data jour- tional Directory of Open Access Journals (DOAJ). nals are published by Elsevier (from which two are pub- The overall number of data papers published by these lished by Academic Press, an imprint of Elsevier, two others data journals is approximately 11,500, with large differ- merged), two by Wiley, one by Springer-Nature, and one by ences, ranging from some papers up to more than 3,500. SAGE. The median number, however is rather low, with ninety- Other data journals are published or hosted by newcom- seven (Figure 4). ers, especially by open access publishers such as Ubiquity In terms of volume, Elsevier’s Data in Brief is by far the Press (three journals), BioMed Central (two journals) most important data journal, followed by Elsevier’s “old” Hindawi, MDPI, Copernicus Publications, Pensoft, or Fac- Atomic Data and Nuclear Data Tables (launched in 1979) and ulty of 1000, by smaller publishing houses like Brill or De Scientific Data, a NatureResearch journal from Springer Na- Gruyter (Sciendo), or by learned societies or university ture. Together, these three journals represent more than presses (AIP, ACS, Wageningen, etc.). half of the data papers published in pure data journals. Most of the data journals are published in three coun- The major business model is OA Gold, mostly with tries, i.e., the United Kingdom, the United States, and The APCs (nineteen) but also without (two). Four journals are Netherlands. The other journals are from Bulgaria, Switzer- hybrid, and only one journal is available through the tradi- land, Germany, and Poland (Figure 3). All are published in tional subscription model (Figure 5). English; only one journal also publishes papers in another In this small sample, there is no “diamond OA journal” language, Dutch (Research Data Journal for the Humanities and without subscription and APCs. In other words, twenty- Social Sciences). five journals (89%) are OA journals or allows OA publish- ing, and in twenty-three journals (82%) authors have to pay 4.3 Business models for OA. All data journals covering the arts, social sciences, and humanities are OA journals, all with APCs. Most of the data journals are “young” products with a short history. Only seven journals were launched before 628 Knowl. Org. 46(2019)No.8 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
Figure 3. Geographical origin of data journals.
Figure 4. Number of data papers per journal (with best estimates).
Knowl. Org. 46(2019)No.8 629 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
Figure 5. Business models.
4.4 Licensing ent manuscript versions of the peer-review completion (post-discussion review of revised submission) will be Twenty-one data journals disseminate data papers with an published if the paper is accepted.10 open license, most often a CC-BY license, sometimes to- gether with a public domain license (CC0) or the more re- 4.6 Structure and length strictive CC-BY-NC-ND or CC-BY-NC-SA licenses (no commercial re-use). Elsevier also proposes its own user li- We already mentioned that it is generally assumed that data cense. Only one journal does not propose an open license papers are short texts, up to four pages. In fact, this is only for the dissemination of the data papers but keeps the full partly true. In this sample, only five journals require short copyright (Journal of Physical and Chemical Reference Data). papers, limited to four to six pages or maximal 3,000 words. Most journals do not limit the length of submitted 4.5 Selection papers or make the usual recommendations (six to ten pages, or maximal 6,000 words). One journal only accepts Except for one title (European Power Watch), all data journals short abstracts (Ecological Archives), while others publish pa- explicitly inform about some kind of formal selection pro- pers well beyond the length of regular papers, up to twenty cedure. Often the information for authors just mention or thirty or even 100 pages, including detailed data descrip- “peer review,” but six describe the selection as a single- tions, illustrations (figures), or data tables like Atomic Data blind review process where the identities of the reviewers and Nuclear Data Tables. On the other hand, data journals are not disclosed to the author(s). One journal (Chemical in the field of the arts, social sciences, and humanities pub- Data Collections) applies a “quick peer review” with focus lish generally shorter data papers. on the data value and potential re-use but does not explain No results, no discussion, no conclusion: usually the who does the peer review and how long it takes. data journal guidelines for authors contain these or similar Five data journals apply some kind of innovative open recommendations, like Elsevier’s Data in Brief, which asks peer review, either as an option (if required) or for all sub- authors to “avoid using words such as ‘study’, ‘results’, and mitted papers. Yet, this term has different meanings: ‘conclusions.’”11 Quite different, the Atomic Data and Nu- clear Data Tables guidelines leave it to the authors whether – the reviewers are suggested (and known) by authors or not to include results, discussion, and conclusion to the (F1000Research); description of the data. – community peer review (Biodiversity Data Journal); Nearly all journals require or suggest a particular struc- – interactive public peer review (Earth System Science Data). ture, and some of them provide a template with manda- tory sections. However, there is no standard structure. In- The last procedure is particularly interesting: all referee and stead of a generally accepted succession of sections, data editor reports, the authors’ response, as well as the differ- papers are made of three constitutive elements, i.e., an in- 630 Knowl. Org. 46(2019)No.8 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data troduction with information about the context and the ra- cussion of these results and an outlook on further re- tionale, a more or less detailed description of the datasets search, very similar to the usual structure of scientific arti- with specifications (sometimes formalized as disciplinary cles and blurring the frontiers between both types of pa- or generic metadata of data, such as the DataCite Metadata pers. Schema or the DDI12), and a section of materials and A last aspect: no invitation or guidelines were found methods, instrumentation, on the production of the data concerning machine-based generation and/or automatic and procedures, sometimes extended to experimental de- processing of data papers. Apparently, the publishers’ plat- signs and calculation (Figure 6). forms do not support automatic ingestion of text files (via The figure presents a core structure with three central FTP of repository metadata or similar) but require manual sections (in blue), with other, optional or peripheral sec- deposits of manuscripts and authorship. Of course, this tions, some of them similar to regular papers (in italics), requirement does not exclude partly or complete machine- others characteristic for data papers, such as: based generation of data papers upstream of the human deposit of manuscripts. – Value and validation: information about the (potential or real) value of the datasets and the quality control 4.7 Metadata and identifiers (validation), like peer review, automatic procedures (technical validation), etc. Two types of metadata must be distinguished regarding – Potential reuse: information about potential usage, data papers, i.e., metadata of the described datasets, and about reuse, and the potential interest for scientists or metadata of the data papers themselves. other users. – Access and availability: information about the address – Metadata of datasets: as mentioned above, some data of datasets (repository, URL) and the availability, in- journals require a detailed and formalized description cluding access and reuse rights and limitations; this part of datasets in a format that is potentially compliant with may include implementation details, about the availabil- metadata. But only a few journals insist on a specific ity of source code and requirements, and about the standard. Two examples: Ecological Archives expects strict availability of supporting data and materials. adhesion to the metadata content standards derived from a set of generic metadata descriptors published by Information about access and availability may also be part the Ecological Society of America (Michener et al. of the appendices, like acknowledgements, references, 1997); the metadata set should be sent to the editor as competing interests, author roles and information, rights a separate text file. Genomics Data requires compliance and permissions, or even peer review comments. with an internal standard for data description with eight As mentioned above, some data journals allow or invite fields.13 Both formats have in common that they are sections about results of data analysis, together with a dis- community-specific, disciplinary metadata standards. A
Figure 6. Sections of a data paper. Knowl. Org. 46(2019)No.8 631 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
third example is quite different, generic, and limited to pects characterise this embeddedness in the new environ- the datasets’ identifiers: Scientific Data requires an ISA- ment: Tab14 metadata text file where the DOI of all datasets are mentioned. Business model: The dominant business model (gold – Metadata of data papers: most journals ask for some OA with APCs) is different from the traditional and still general and usual information, compliant with the Dub- prevailing serials landscape, and it appears already com- lin Core format, such as author, organisation, title pliant with the requirements of the new plan S.16 etc. F1000Research recommends XML Schema, Xlink, Reuse rights: most data journals allow publishing with MathML, or the NLM Journal publishing DTD (JATS15). an open license, often with generous reuse and remixing rights (e.g., CC-BY license and/or CC0 waiver). Twenty-six journals publish the data papers with a DOI Findability: the editorial model of data journals requires (93%), and five also include the author identifier ORCID standard identifiers for the datasets, e.g., DataCite’s DOI, (18%). Also, most of them recommend if not require a to guarantee (and increase) the findability of datasets; standard identifier (DOI) or at least a stable address for they also attribute DOIs to their own data papers, creat- the described datasets. ing a kind of cross-linked DOI system between data pa- pers and datasets. 4.8 Linking Interconnectedness: perhaps the most relevant aspect is the integration of data journals and papers in a com- All data papers provide information about the availability plex structure of open access journal platforms and data of the described datasets, mostly together with an address repositories, academic communities, research projects, (URL), but they do it in different ways: conferences, etc. Interconnectedness requires interoper- ability between platforms and infrastructures but is more – usually in a special section of the paper with a statement than technology, formats, and standards, insofar it means on data access and availability, new ways of doing science, including research manage- – in an appendix that contains a declaration with data ment, research environment, workflows, etc. availability and address, – in the abstract, A fifth aspect, i.e., evaluation and selection, is already visi- – as part of the metadata. ble but still in transition and not dominant. Data journals replace the usual evaluation and selection procedure (dou- Some papers contain downloadable data; others require ble-blind peer review) by partly open single-blind peer re- that the described datasets should be deposited in one or view and, for already one out of five journals, by some a shortlist of recommended repositories. kind of open peer review, including innovative community peer review and interactive public peer review. They can 5.0 Discussion also contribute to the assessment of data value through the follow-up of citations (Belter 2014). 5.1 A new ecosystem 5.2 FAIR principles and beyond Compared to former studies, the number of data journals and papers appears to increase slowly on a low level. Garcia- Most data journals have never been produced as traditional Garcia et al. (2015) identified twenty pure data journals; four serials but are a pure (and young) product of this new eco- years later, our sample consists of twenty-eight data journals system of open access, open (and big) data, and new forms and not all are still active and even pure (see below). Twenty- of selection and dissemination. This makes them particu- eight journals represent less than 0.01% of the academic larly different from other academic journals. And this and scholarly serials (source: Scopus). The arts, social sci- makes them also particularly interesting for the require- ences, and humanities are nearly missing (two journals in ments of the so-called “FAIR Guiding Principles for sci- 2015; four in 2019). The number of data papers progressed entific data management and stewardship” (Wilkinson et at a faster pace, from 846 in 2013 (Candela et al. 2015) to an al. 2016). Their data papers contribute to these principles estimated number of 11,500 data papers in 2019. Yet, this in different ways, in order to improve the findability, acces- volume represents roughly 0.4% of the overall number of sibility, interoperability, and reuse of research data, e.g.17: articles published in 2017 (source: Scopus). Also, the interest of data papers and journals is not their – Findable volume but the fact that they clearly are a product of the – F2. Data are described with rich metadata: data pa- emerging ecosystem of data-driven open science. Four as- pers enrich existing metadata of datasets. 632 Knowl. Org. 46(2019)No.8 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
– F4. (Meta)data are registered or indexed in a search- able resource: the enriched metadata are registered, Standardization: the quality of data papers depends indexed, and preserved on the data journal platform. for much on the quality of the metadata of the under- – Accessible lying datasets; and this means, on controlled terminol- – A2. Metadata are accessible, even when the data are ogies, on standard formats, well-defined elements, etc. no longer available: the accessibility of metadata One example is the International Geo Sample Number published via data papers does not depend on the (IGSN) designed to provide an unambiguous globally datasets’ accessibility in a data repository. unique persistent identifier (PID) for physical samples – Interoperable (specimens) and to facilitate the location, identification, – I1. (Meta)data use a formal, accessible, shared, and and citation of physical samples used in research.19 The broadly applicable language for knowledge represen- development of data papers and data journals should tation: at least some of the data journals insist on the (will) be accompanied by further work on standards, by application of formal, standard language (vocabular- academic communities, publishers, information profes- ies) for the description of datasets. As a minimum, sionals, and knowledge practitioners. they reproduce the data repositories’ own formal da- Specialisation: to be relevant and useful, metadata taset representation. standards should be as compliant as possible with the – I3. (Meta)data include qualified references to other specific requirements and features of scientific commu- (meta)data: data papers can (and usually do) provide nities, disciplines, methods, tools, and equipment. This links to other related resources, e.g., research papers, specialisation, however, tends to limit their interoperabil- institutional affiliations, similar or related datasets, ity between different domains, infrastructures, infor- etc. mation systems, and their interest and usefulness for in- – Reusable terdisciplinary research, discovery tools, etc. One solu- – R1.1. (Meta)data are released with a clear and acces- tion to this problem could be described by “as specific as sible data usage license: as mentioned above, most possible, as generic as necessary,” an approach that would data papers are published with an open license; apply a kind of ad-hoc-compromise for each particular whenever the data paper is derived from the original situation, resulting in many different formats more or metadata, this license may depend on the reposi- less specific, and more or less generic. Another, perhaps tory’s initial licensing and reuse rights. more realistic approach would be to accept (and support) – R1.2. (Meta)data are associated with detailed prove- two (or more) different standards for each dataset and nance: one of the main functions of data papers is each data paper, one generic (like, for instance, the to provide detailed knowledge about where the data DataCite metadata schema), the other specific, depend- came from, who to cite, who generated or collected ing on the particular domain, method, tool, etc. it, and how has it been processed (workflow). 5.3 Blurred boundaries Along with metadata, data papers contribute to the com- pliance with FAIR principles, in particular to the two prin- The specific identity of data papers is mainly defined in ciples of findability and reusability, insofar they help peo- opposition with regular research papers (see for instance ple (and machines18) find datasets and inform about the Penev et al. 2012). The reality is different. The empirical provenance and reuse rights. Additionally, data papers con- data of our survey provides evidence that despite a general tribute to another aspect, beyond the FAIR principles, i.e., definition of data papers and journals, there is a lot of di- the evaluation of the datasets’ quality and value. vergence and heterogeneity that can be described on four In the context of open science, metadata has been con- levels. sidered fuel for economy (Neuroth et al. 2013). As a new vector of communication of metadata on research data, 1. Data journals also accept other articles. Our sur- data papers can be defined as a kind of pipeline for this vey put the focus on a limited number of academic and fuel. Yet, as they also add value to metadata, through con- scholarly journals indexed by databases or directories as textual information, evaluation, new identifiers, etc., they “pure” data journals. Yet, even in this sample some data are not only pipelines but also refineries, more or less spe- journals publish regular research articles, reviews, short cialised, more or less standardized. To stay with the fuel communications, or comments along with data papers, metaphor, data papers are a new infrastructure of refine- such as Data from MDPI and Earth System Science Data ment and dissemination of the metadata fuel. from Copernicus. Regarding knowledge organization, two aspects require 2. Data papers are published in other journals. As attention and further investigation: mentioned above, one limitation of our survey is the fo- Knowl. Org. 46(2019)No.8 633 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
cus on supposedly “pure” data journals. However, an in- These two examples of a new kind of paper are quite creasing number of academic and scholarly journals ac- different, yet they have in common that they are both cept data papers along with regular research papers, usu- linked to research datasets and, above all, to the dissem- ally in a specific section. Pensoft for instance publishes ination and reuse of research data, which is their main thirty-seven journals, including one data journal and six- purpose. teen other journals accepting data papers. The French Agricultural Research Centre for International Develop- The boundaries between data papers and data journals and ment (CIRAD) produced a list with fifty-four academic other categories of scientific communication are partly journals accepting data papers relevant for agricultural blurred, not only due to a lack of reference definitions but science, including the mega-journal PLoS One.20 It is also due to a large diversity of publishing practices. This quite impossible to make an estimation of the real num- may have at least three explanations: ber of such “mixed” data journals and their data papers. Pensoft’s Research Ideas and Outcomes for instance is part of – The publishing of data papers is still in transition. It these new “mixed” data journals but published up to now took some decades to develop and accept the IMRAD only one data paper, which was in biosciences. format as a standard format of scientific article pub- 3. Data papers are more than simple data papers. lishing.23 The heterogeneous character and blurred Even a superficial analysis of data papers reveals that boundaries of data papers may reflect the emergence one part of articles labeled as “data papers” do not only of a young and new, still not well-defined form of sci- describe datasets but add data analysis and discussion entific communication. of results. Atomic Data and Nuclear Data Tables, Dataset – The described proximity with research communities, Papers in Science, and Open Archaeology are three “pure” the “embeddedness” in an ecosystem defined by disci- data journals that explicitly accept data papers with re- plines, materials, methodologies, tools, etc., contributes sults and discussion of results. This means that a (un- to the heterogeneity of data journals and papers. Data known) part of data papers, in fact, are more than sim- papers necessarily depend on the community-specific ple data papers stricto sensu, because they communicate way of how data is produced, collected, processed, pre- results of data analysis. served, reused, and it seems quite natural that they will 4. There are other emerging types of articles, sim- reflect the diversity of this environment. Perhaps, fuzz- ilar to but not identical with data papers. “Pure” and iness is a core element of the data paper category. “mixed” data journals are open for other categories of – One part of the new OA journals announces an inclu- articles that are neither traditional journal items (re- sive editorial policy. Instead of a selective approach and search articles, reviews, comments, etc.) nor data pa- guidelines with explicit limitations, they invite submis- pers. Sometimes the difference may be a question of sion of all kind of papers; a strategy somewhere be- terminology. For instance, F1000Research accepts “brief tween predatory publishing and big data principles descriptions of scientific datasets that promote the po- based on volume and variety rather than on quality and tential reuse of research data and include details of why trustworthiness. and how the data were created” called “data notes”21— in other words, data papers. But there are other exam- 5.4 Who is writing? Who is reading? ples (see also the listed terms in Candela et al. 2015): a. Data services paper: “papers on data services, and Some of the underlying questions of this study were about papers which support and inform data publishing the production and use of data papers. How are they writ- best practices (including) the development of sys- ten, and are they really “written?” Which is the (potential tems, techniques or tools that enable data analysis, and real) part of automatic or semi-automatic production, data visualisation, data collection and data sharing and which is the part of human added value? In fact, are (and) processes and procedures used in the develop- data papers written by machines and to be read by ma- ment of datasets” (Geoscience Data Journal). chines? b. Meta or overlay articles: “Descriptions of online The answer to these questions is neither yes nor no. As simulation, database, and other experiments, part- mentioned above, data papers can be at least research data nering with digital repositories on ‘meta articles’ or available in data repositories such as Dataverse or others (see ‘overlay articles’, which link to and allow visualisa- the Pensoft workflow, Chavan and Penev 2011). The tech- tion of the data, thereby adding an entirely new di- nology is there. Recently, the French National Institute for mension to the communication and exchange of Agricultural Research (INRA) updated their Dataverse- data research results and educational materials” based repository including an online tool that partly gener- (Data Science Journal).22 ated “by machines,” i.e., through the exploitation and trans- 634 Knowl. Org. 46(2019)No.8 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data formation of metadata on researchers can use to generate ically generated data papers. Rich and less standardized and data papers from the deposits’ DOI, in an open text format coded textual discussion, for instance, is probably more compliant with INRA’s own data journal or with Elsevier’s aimed at human readers. This of course does not exclude Data in Brief.24 the potential of data papers for automatic exploitation Both examples, the Pensoft workflow as well as the with tools of text and data mining (artificial intelligence). INRA tool, reveal the potential of automatic generation of Similar to the generation (writing), this potential depends data papers but also its requirements and limits. Automatic on the standardization of data papers, including careful generation of data papers requires a high degree of stand- coding, and their own metadata, i.e., standardized and well ardization and interoperability between data repositories, controlled formats and terminology. Probably, the fast de- text processing tools and journal platforms, especially re- velopment of artificial intelligence will facilitate the auto- garding metadata formats and identifiers. Our study was not matic production as well as the automatic exploitation of about metadata formats of data repositories and about their data papers and their metadata. However, so far, we did degree of standardization. But our study reveals a lack of not identify any study about this potential, which for the standardization on the other side, i.e., the journal platforms. moment apparently remains theoretical. Paradoxically, this may be an opportunity for automatic gen- eration and ingestion of data papers; yet it will not be helpful 5.5 Data? Information? Knowledge? for machine-based exploitation of data papers. The limits of automatic generation of data papers are Finally, what is the informational status of data papers, twofold. On the one hand, journal platforms still and al- compared with the DIKW model of information sciences ways require authorship, i.e., intellectual property and in- (Rowley 2007)? What do they carry: data, information, stitutional affiliation. They do not accept automatic sub- knowledge, or wisdom? Following the usual definitions, mission of machine-produced data papers. On the other the answer seems easy: insofar data papers provide de- hand, the format of data papers requires rich contextual scription of data, and insofar information is inferred from information that may not be part of the datasets’ metadata data and contained in descriptions (Rowley and Hartley and must be added by the researchers or data officers. Can- 2008), data papers correspond to the second level of the dela et al. (2015) mention that the metadata is usually se- DIKW pyramid, i.e., information (Figure 7). They are not lected by both the data journal editor (for the data paper) knowledge but contribute to the generation of knowledge. and the data archive manager (for the dataset), which “of- Also, the main purpose of data papers—to facilitate the ten results in proprietary, ad-hoc-solutions.” Relevant for findability and the reusability of research data—is similar our question is the human contribution (selection) and the to another general aspect of information, i.e., its immedi- resulting diversity and specificity. ate usefulness for decisions or actions. Candela et al. (2015) also insist on the distinction be- This characteristic of data papers is one major differ- tween metadata of datasets, metadata of data papers, and ence with regular research articles, which are expected to data papers themselves. Metadata25 are made for machines, provide more than simple descriptions of facts (data), i.e., and the main purpose of FAIR principles is to improve insight, understanding, interpretations, hypotheses, etc. machine readability and transfer of research data. Data pa- However, as mentioned above, the boundaries are partly pers are part of this ecosystem, and they contribute to the blurred and some data papers do more than carrying in- automatic processing of research data and related formation about data, in particular when they include sec- metadata. However, the state of the art and our empirical results (still) reveal human added value, i.e., enhancement of the information produced by metadata, such as poten- tial reuse (value), related datasets and research, and other contextual information useful for the understanding of the described data. But as mentioned above, this can also in- clude more traditional content, like results of data analysis and rich textual discussion of data and results. Another “human added value” is the intellectual responsibility and property of the data papers, which are always attributed to people (authors) not to machines. Instead of machine gen- erated data papers, we should speak of “machine- (or re- pository-) assisted writing of data papers.” So, are data papers written for machines? Penev et al. Figure 7. Data papers and the DIKW pyramid. (2012) insist on the “human-readability” even of automat- Knowl. Org. 46(2019)No.8 635 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data tions with data analysis results and discussions. So at least 2. However, the boundaries with other categories of aca- partly, data papers also convey knowledge, even if this is demic publishing are partly blurred, especially with reg- not part of their core function. ular research papers. Downside of the pyramid, the boundary to the data level 3. Data papers are (can be) generated automatically and seems equally blurred. Because, as described above, data pa- are potentially machine-readable; yet, the human con- pers do not only provide information about data but can be tribution (still) appears vital in terms of intellectual exploited as raw data on their own, generating information property and richness of content. about research projects, scientific cooperation, etc. This 4. Data papers are essentially information, i.e., description means that data papers are also partly data. of data (as defined by the DIKW model) but also partly For both reasons, data papers do not just improve the contribute to the generation of knowledge and data on referencing of datasets on repositories but fulfil other roles. its own. Their particular information profile can be described in terms of library science, as an original integration (or merg- As to the two camps, human generated vs. machine gener- ing) of writing, cataloguing, and indexing, facing major chal- ated, if a data paper is created by a human—whether or lenges like standards and terminology. Perhaps data papers not machine aided, one can speak of knowledge organiza- are a kind of new boundary object (Star and Griesemer tion. However, if the data paper is solely machine gener- 1989) on the frontline between academic publishing and re- ated it is difficult to attribute this to knowledge organiza- search. Our analysis confirms the statement that data papers tion (excluding any reference to artificial intelligence). The are like traditional research papers in some aspects but very latter is more aligned with automated indexing, catalogu- different in other respects (Smith 2012). Perhaps data papers ing, and the like. are not (only) part of academic publishing but should (also) In relation to the DIKW pyramid, data papers appear be considered and assessed as part of research data practice. between the levels of information and knowledge given that for some people they are not knowledge but only con- 6.0 Conclusion tribute to the generation of knowledge. However, if one looks at the metadata fields that encom- Data papers have been defined as scholarly journal publi- pass a full blown data paper—such as the explicit roles of cations whose primary purpose is to describe research data the researchers/authors, the research methods applied, the (facts about data). Yet, the literature overview shows that description of the data, its reusability as well as its limita- there is a lack of a generally accepted reference definition tions, then one may conclude that the data paper provides a of data papers. Likewise, few conceptual studies and em- fuller understanding of the data/dataset. In itself, the data pirical evidence are provided. Also, up to now, the success paper provides a best practice in knowledge organization— of data papers appears of minor importance and limited if not an example of knowledge generation. to STM disciplines, primarily in the life sciences. Part of the new ecosystem of open and data-driven sci- Our survey provides more insights about the environ- ence, data papers and data journals are an interesting and ment of data papers, i.e., disciplines, publishers, and busi- relevant object for the assessment and understanding of the ness models, and about their structure, length, formats, transition of the former system of academic publishing. metadata, and licensing. Core elements of data papers are This means that the quality and the usefulness of data pa- the data description and methods and materials; depending pers partly depend on external variables, e.g., the metadata on the data journal’s policy, other sections are requested or standards of data repositories, their trustworthiness in optional, such as value and validation, potential reuse, ac- terms of data quality but also long-term preservation (certi- cess and availability, and even results of data analyses and fication), etc. Therefore, as mentioned above, quality control discussion of results. of data papers (i.e., some kind of peer review) always im- The discussion section of this study highlights five ma- plies some kind of quality control or evaluation of the da- jor aspects of data papers: tasets themselves and their respective repositories. Based on our empirical results and former studies, we 1. Data papers are a product of the emerging ecosystem would suggest the following definition of data papers, of data-driven open science. keeping in mind the transitional and necessarily provi- They contribute to the FAIR principles for research sional character of each conceptual attempt: data management, in particular findability and reusabil- ity, and add in some degree to the evaluation of the Data papers are authored, peer reviewed and citable quality and value of the data. articles in academic or scholarly journals, whose main content is a description of published research datasets, along with contextual information about 636 Knowl. Org. 46(2019)No.8 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
the production and the acquisition of the datasets, 2. Source: data from Dimensions https://www.dimen with the purpose to facilitate the findability, availa- sions.ai/ bility, and reuse of research data; they are part of the 3. Source: Global Biodiversity Information Facility research data management and crosslinked to data https://www.gbif.org/data-papers repositories. 4. See for instance http://researchdata.cab.unipd.it/122/ 5. Article processing charges: the fee authors or their in- This definition may not cover all different variants of data stitutions have to pay (after the acceptation of their papers but will be helpful for a better understanding of papers) to some publishers to be published immedi- what we called “blurred boundaries” and for further inves- ately in open access. The amount of APC is varying tigation. between publishers and journals; the average amount At this stage, a couple of questions remain open; in par- research institutions pay per article is about 2,000 eu- ticular, the following topics should be addressed: ros (see OpenAPC https://treemaps.intact-project. org/apcdata/openapc/). – Monitoring: how can the indexing of data papers be im- 6. Chemical Data Collections, see https://www.else- proved in order to facilitate their identification and fol- vier.com/journals/chemical-data-collections/2405-83 low-up (search engines, databases, data repositories, 00/guide-for-authors journal platforms)? 7. FOSTER portal, see https://www.fosteropenscience. – Business models: what is the risk of predatory publish- eu/foster-taxonomy/open-data-journals ing of data journals and data papers? Is it different from 8. forschungsdaten.org, see https://www.forschungsda predatory publishing of regular research papers? ten.org/ – Disciplines: are data papers as relevant in the arts, social 9. Both in the field of agronomy: INRA https://www6. sciences, and humanities as in life sciences, chemistry, inra.fr/datapartage/Partager-Publier/Publier-un-Data- etc.? Should their data papers be published in large and Paper and CIRAD http://ou-publier.cirad.fr/formu multidisciplinary data journals, together with STM, or laire.php should they have their own data journals? 10. See https://www.earth-system-science-data.net/peer_ – Ecosystem: more case studies are needed on specific review/interactive_review_process.html links between research data management, academic 11. See https://www.journals.elsevier.com/data-in-brief/ publishing, and the production and dissemination of about-data-in-brief/how-to-submit-a-data-in-brief-ar data papers in a given environment and community ticle (equipment, discipline, structure, etc.). 12. DataCite https://schema.datacite.org/ and Data Doc – Evaluation: our study did not assess whether (and how) umentation Initiative https://www.ddialliance.org/ scholars get credit for publishing data papers. This, 13. These eight fields are: organism/cell line/tissue; sex; however, will be a key factor for the future development sequencer or array type; data format; experimental fac- of data papers. tors; experimental features; consent; sample source lo- cation. Garcia-Garcia et al. (2015) wondered if data journals will re- 14. ISA tools https://isa-tools.org/ main part of the research ecosystem or not. Perhaps they 15. Journal Publishing Tag Set https://jats.nlm.nih.gov/ will not. However, it seems probable that the number of publishing/ data papers will continue to grow and gain importance, per- 16. The plan S gives preference to immediate open access in haps (probably) not via data journals but via increasing hy- 100% OA journals, see https://www.coalition-s.org/ bridization of research journals and journal platforms, and 17. The description and numbering of the principles fol- perhaps even through the merging of journal and data plat- low the GO FAIR list at https://www.go-fair.org/fair- forms. In any case, on the boundary between research data principles/ management and academic publishing, data papers will con- 18. The FAIR principles have been initially designed for tinue to provide a highly relevant object for library and in- automatic data processing. formation science, especially for the further assessment of 19. ISGN http://www.igsn.org/ the development of academic publishing and knowledge or- 20. CIRAD, see http://ou-publier.cirad.fr/index.php ganization in the field of scientific research. 21. F1000Research, see https://f1000research.com/for-au thors/article-guidelines/data-notes Notes 22. Data Science Journal, see https://datascience.codata. org/about/ 1. GreyNet International, Amsterdam; see http://www. 23. IMRAD is a common organizational structure of sci- greynet.org entific writing and the usual format of papers on orig- Knowl. Org. 46(2019)No.8 637 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
inal research published as articles in scientific journals, De Waard, Anita. 2010. “The Future of the Journal? Inte- in particular in empirical sciences but also in other dis- grating Research Data with Scientific Discourse.” Logos ciplines. It stands for “introduction, methods, results 21, no. 1-2: 7-11. doi:10.1163/095796510X546878 and discussion/conclusion.” For more details and ref- Farace, Dominic J., Jerry Frantzen and Plato L. Smith. erences, see https://en.wikipedia.org/wiki/IMRAD 2018. “Data Papers are Witness to Trusted Resources 24. INRA, see https://data.inra.fr/datapartage-datapapers- in Grey Literature: A Project Use Case.” Grey Journal 14, web/ and https://dataverse.org/blog/data-inra no. 1: 31–36. 25. Metadata considered in the strict sense of the term, i.e., Friedman, Rachel, Stéphanie Psaki and Jeffrey B. Bingen- digital data on other digital data. heimer. 2017. “Announcing a New Journal Section: Data Papers.” Studies in Family Planning 48: 291-2. doi:10. Data availability 1111/sifp.12032 Garcia-Garcia, Alicia, Alexandre Lopez Borrul and Fer- The CSV table of the underlying dataset is available in the nanda Peset. 2015. “Data journals: eclosión de nuevas Dutch repository DANS, at the following address: https:// revistas especializadas en datos.” El profesional de la infor- doi.org/10.17026/dans-zk3-jkyb mación 24: 845-54. doi:10.3145/epi.2015.nov.17 Huang, Xiaolei, Bradford A. Hawkins and Gexia Qiao. References 2013. “Biodiversity Data Sharing: Will Peer-Reviewed Data Papers Work?” BioScience 63: 5-6. doi:10.1525/ Belter, Christofer W. 2014. “Measuring the Value of Re- bio.2013.63.1.2 search Data: A Citation Analysis of Oceanographic Le Deuff, Olivier. 2018. “Une nouvelle rubrique pour la Data Sets.” PLoS One 9, no. 3: e92590. doi:10.1371/ RFSIC : Le data paper.” Revue française des sciences de l’in- journal.pone.0092590 formation et de la communication 15. http://journals.open Bordelon, Dominic, Uta Grothkopf, Sylvia Meakins and edition.org/rfsic/5275 Michael Sterzik. 2016. “Trends and Developments in Li, Kai, Jane Greenberg and Jullian Dunic. 2019. “Data VLT Data Papers as Seen Through Telbib.” In Observa- Objects and Documenting Scientific Processes: An tory Operations: Strategies, Processes, and Systems VI: 27 June– Analysis of Data Events in Biodiversity Data Papers.” 1 July 2016 Edinburgh, United Kingdom, ed. Alison B. Peck, Journal of the Association for Information Science and Technol- Robert L. Seaman and Chris R. Benn. Proceedings of ogy early view. doi:10.1002/asi.24226 SPIE 9910. Bellingham, WA: SPIE. doi:10.1117/ MESRI (Ministère de l’Enseignement Supérieur, de la Re- 12.2231697 cherche et de l’Innovation). 2018. “National Plan for Callaghan, Sarah, Steve Donegan, Sam Pepler et al. 2012. Open Science.” https://libereurope.eu/wp-content/ “Making Data a First-Class Scientific Output: Data Ci- uploads/2018/07/SO_A4_2018_05-EN_print.pdf tation and Publication by NERC's Environmental Data Michener, William K., James W. Brunt John J. Helly, Centres.” International Journal of Digital Curation 7: 107- Thomas B. Kirchner and Susan G. Stafford. 1997. 13. doi:10.2218/ijdc.v7i1.218 “Nongeospatial Metadata for the Ecological Sciences.” Candela, Leonardo, Donatella Castelli, Paolo Manghi and Ecological Archives 7, no. 1: 330-42. doi:10.2307/2269427 Alice Tani. 2015. “Data Journals: A Survey.” Journal of Neuroth, Heike, Stefan Strathmann, Achim Osswald and the Association for Information Science and Technology 66: Jens Ludwig, eds. 2013. Digital Curation of Research Data. 1747-62. doi:10.1002/asi.23358 Glückstadt: Werner Hülsbusch. Chavan, Wishwas and Lyubomir Penev. 2011. “The Data Newman, Paul and Peter Corke. 2009. “Data Papers – Peer Paper: A Mechanism to Incentivize Data Publishing in Reviewed Publication of High Quality Data Sets.” The Biodiversity Science.” BMC Bioinformatics 12, S2. doi: International Journal of Robotics Research 28: 587. doi:10. 10.1186/1471-2105-12-S15-S2 1177/0278364909104283 Costello, Mark J., William K. Michener, Mark Gahegan, Pärtel, Meelis 2006. “Data Availability for Macroecology: Zhi-Quiang Zhang and Philipp E. Bourne. 2013. “Bio- How to Get More out of Regular Ecological Papers.” diversity Data Should Be Published, Cited, and Peer Re- Acta Oecologica 30: 97-9. doi:10.1016/j.actao.2006.02.002 viewed.” Trends in Ecology & Evolution 28: 454-61. doi: Penev, Lyubomir, Wishwas Chavan, Teodor Georgiev and 10.1016/j.tree.2013.05.002 Pavel Stoev. 2012. “Data Papers as Incentives for Open- Davis, Grace H., Eric Payne and Andrew Sih. 2015. “Com- ing Biodiversity Data: One Year of Experience and Per- mentary: Four Ways in Which Data-Free Papers on An- spectives for the Future.” https://pensoft.net/img/ imal Personality Fail to Be Impactful.” Frontiers in Ecology upl/file/DataPaperPoster.pdf and Evolution 3, no. 102: 1-3. doi:10.3389/fevo.2015. 00102 638 Knowl. Org. 46(2019)No.8 J. Schöpfel, D. Farace, H. Prost, A. Zane. Data Papers as a New Form of Knowledge Organization in the Field of Research Data
Reymonet, Nathalie. 2017. “Améliorer l'exposition des Smith, Mackenzie. 2012. “Data Papers in the Network données de la recherche : La publication de data pa- Era.” In Something's Gotta Give: Charleston Conference Pro- pers.” https://archivesic.ccsd.cnrs.fr/sic_01427978/ ceedings, 2011, ed. Beth R. Bernhardt, Leah H. Hinds and Rowley, Jennifer. 2007. “The Wisdom Hierarchy: Repre- Katina P. Strauch. West Lafayette, IN: Against the sentations of the DIKW Hierarchy.” Journal of Infor- Grain Press, 13-20. doi:10.5703/1288284314871 mation Science 33: 163–80. doi:10.1177/0165551506070 Star, Susan Leigh and James R. Grisemer. 1989. “Institu- 706 tional Ecology, `Translations’ and Boundary Objects: Rowley, Jennifer and Richard Hartley. 2008. Organizing Amateurs and Professionals in Berkeley’s Museum of Knowledge: An Introduction to Managing Access to Information. Vertebrate Zoology, 1907-39.” Social Studies of Science 19: Aldershot: Ashgate. 387–420. doi:10.1177/030631289019003001 Senderov, Viktor, Teodor Georgiev and Lyubomir Penev. Wilkinson, Mark D., et al. 2016. “The FAIR Guiding Prin- 2016. “Online Direct Import of Specimen Records into ciples for Scientific Data Management and Steward- Manuscripts and Automatic Creation of Data Papers ship.” Scientific Data 3: 160018. doi:10.1038/sdata. from Biological Databases.” Research Ideas and Outcomes 2: 2016.18 e10617+. doi:10.3897/rio.2.e10617
Knowl. Org. 46(2019)No.8 639 P. Gomes and M. Guiomar da Cunha Frota. Knowledge Organization from a Social Perspective
Knowledge Organization from a Social Perspective: Thesauri and the Commitment to Cultural Diversity † Pablo Gomes* and Maria Guiomar da Cunha Frota** Universidade Federal de Minas Gerais, Escola de Ciência da Informação Av. Antônio Carlos, 6627, Pampulha, Belo Horizonte, Minas Gerais 31270-901, Brazil, *
Pablo Gomes is a doctoral student with a master’s degree in information science at the Federal University of Minas Gerais. He graduated in librarianship at the Federal University of Ceará and is Professor at the Federal Institute of Education, Science and Technology of Maranhão.
Maria Guiomar da Cunha Frota has a PhD in sociology at the University Institute of Research of Rio de Janeiro, and a master’s degree in sociology at the Federal University of Minas Gerais. She graduated in history at the Pontifical Catholic University of Minas Gerais and is Professor and coordinator for the Graduate Program in Information Science at the School of Information Science of the Federal University of Minas Gerais.
Gomes, Pablo and Maria Guiomar da Cunha Frota. 2019. “Knowledge Organization from a Social Perspective: Thesauri and the Commitment to Cultural Diversity.” Knowledge Organization 46(8): 639-646. 19 references. DOI:10.5771/0943-7444-2019-8-639.
Abstract: Knowledge organization systems can have linguistic and conceptual formations of social oppression and exclusion. It is information science’s role to be vigilant in perpetuating seditious discourses which end up reaffirming offenses, prejudices and humiliations to certain groups of people, especially those labeled as margin- alized, that is, who are not part of the dominant group holding social power. In the quest for this diversity, this study reviews the literature of the area on how thesauri can become more inclusive and on the role of semantic warrant, specific to the philosophical, literary and cultural warrant. This research highlights the need to review thesaurus construction models so that they can be more open and inclusive to the cultural diversity of today’s society, formed by social actors who claim their spaces and representations. To this end, guidelines are suggested for the construction of thesauri procedures that allow cultural warrant receptivity.
Received: 30 September 2019; Revised: 28 October 2019, 6 November 2019; Accepted: 15 November 2019
Keywords: cultural diversity, cultural warrant, thesaurus construction, information
† Presented at ISKO-Brazil 2019: Organização do conhecimento responsável: Promovendo sociedades democráticas e inclusivas (Respon- sible Knowledge Organization: Fostering Democratic and Inclusive Societies), at Umarizal (Belém), Pará, Brazil, September 2-3, 2019.
1.0 Introduction Thesauri have long become tools for librarians and infor- mation professionals to represent information, dating How does one build a thesaurus for some knowledge areas from the 1950s, thus for almost seventy years. During of human and social sciences? The answer to this question these years, when thesauri contributed to facilitating both may seem quite simple: following thesaurus construction representation and information retrieval, little concern was models. However, other questions emerge as we take a given to the exclusionary relationships resulting from the closer look at such models: 1) Do they account for the absence of a more critical analysis regarding the represen- complexity of areas that change so constantly in terms of tation forms of diversity. themes and issues?; 2) Do they cover and cope with issues If we consider a language, in this case documentary lan- that are also the subject of disputes in the political and so- guage (DL), we need to understand that the use of certain cial field?; and 3) Do these models allow the inclusion of terms in certain settings is relevant in perpetuating forms cultural diversity addressed by these sciences? The answer of social oppression, understood in the context of this re- to these questions is that to embrace the complexity sur- search as offensive, biased and humiliating structures to rounding cultural diversity representation requires critical certain groups of people, especially those identified as mi- analysis and the adaptation or creation of new models. norities such as women, black people, LGBT+ and others. 640 Knowl. Org. 46(2019)No.8 P. Gomes and M. Guiomar da Cunha Frota. Knowledge Organization from a Social Perspective
To cope with the terminological and conceptual part of sity? The general objective was to propose guidelines thesauri, warrants for maintaining semantic quality have within thesaurus construction models that would open been included in thesaurus-building processes. Some have cultural diversity through cultural warrant and be able to been highlighted, such as the philosophical, literary and cope with the complexity of these sciences. use warrants. Others, such as the cultural warrant, have failed to find a place in the practical field on the large-scale. 2.0 Thesaurus and cultural diversity Such warrants are responsible for the content these the- sauri will have associated with them, whether for diversity We have lived moments of social construction when mi- or for maintaining discourses of social exclusion. norities, those left apart from the structure of society, By seeking the semantic reduction of a term and con- claim their spaces and their voices. Thus, rethinking all so- cept in a given area of knowledge, considering the special- cial elements is fundamental to give due space to all those ized language (SL) of the field, thesauri need to be aware who are part and contribute to the construction of society. of the social interlacing certain linguistic formations may This process implies the inclusion of cultural diversity in reflect on the collection and user communities. Thus, the all sectors: in labor, academia, politics, science, infor- adoption of warrants ensures inclusion, as the case of cul- mation, among others. It is crucial that all instances of in- tural warrant. formation science (IS) studies be concerned with the issue It is already seen in the knowledge organization literature of social inclusion of minorities to be able to achieve cul- that DL such as Universal Decimal Classification (UDC) tural diversity. Knowledge and information organization and Dewey Decimal Classification (DDC) have formations with should seek to represent, through its instruments, tools notations that promote, for example, religious intolerance, and information products, the various forms of cultural as the case presented by Trivelato and Moura (2017) when diversity, so that greater identification of users, institu- analyzing the UDC. This research, in addition to many oth- tions, information systems or services is achieved. ers, leads us to resume the discussion regarding the neutral- Thesauri are widely used information organization ity of the information action. Thesauri do not deviate from tools by professionals such as librarians, archivists, mu- the classification tools pointed out above, as they are simpli- seum professionals and others. For a long time, these in- fied representations of the conceptual structure of an area struments have been known to facilitate information rep- through a hierarchical linguistic structure. In their construc- resentation and retrieval. However, advances in ICT have tion processes, they end up being judged by those who par- led to greater global cultural interaction at an unimaginable ticipated in making decisions implying partial interpretations speed for over sixty years, and thesauri have failed to keep and points of view. Guimarães and Pinho (2007) state that up with the expectation of indexing information encom- representation systems reflect pre-established standards set passing global cultural diversity. out by the creator of a knowledge organization tool. There- One of the elements that eventually caused this lack of fore, it is utopian to speak of neutral information systems, representativeness in thesauri was its characteristic of bear- instruments, or actions, as the worldviews of the creators of ing a single voice, that is, a single discourse, usually the sci- such elements will be implicitly implicated in the process. entific-academic one. This characteristic provided what Cin- Thinking of the transition from neutrality to ethical action tra et al. (2002) and Dodebei (2002) named as “economy;” is crucial, that is, changing the mindset that there is neutrality not in the sense of area of knowledge but regarding an in information activity to an ethical information activity con- economy of meaning, as it reduces the scope of a term cerned with the impact that each action may reflect. They within a context. are in line with Barité (2011) by understanding that However, opening up the possibility of including several knowledge organization systems (KOSs) entrenched with voices in the same thesaurus does not mean removing their the worldviews of one single culture are excluded from cul- contextual limitation characteristic. To think about the pos- tures other than the one privileged in the construction of sibility of the thesauri to accommodate the views and values the system. Their critique of the construction conducting a of more than one culture is in line with the studies of cul- KOS, either by the creator’s interpretation of the instrument tural hospitality (Beghtol 2002). This is the ability of a KOS or by the worldview a single culture can offer, can produce to incorporate more than one worldview into the process of social oppression by maintaining and continuing dominant knowledge representation. The author seeks to include the discourses, which may be offensive to certain groups of various cultures as a preponderant factor in the fulfillment people. of an ethical posture as a discussion point in KOS. In this In this context, this research was guided by the ques- way, cultural hospitality would bring the thesauri a multitude tion: how can thesaurus construction procedures be better of voices. suited to areas such as the humanities and social sciences Culture “leads us to a complex conceptual universe, through cultural warrant and by expanding cultural diver- made up of innumerable theoretical-ideological strands that Knowl. Org. 46(2019)No.8 641 P. Gomes and M. Guiomar da Cunha Frota. Knowledge Organization from a Social Perspective reflect the different views on its conception and existing di- translation). Following this line, Zeng (apud Carlan 2010, 31, mensions” (Boccato and Biscalchin 2014, 39, our transla- our translation) indicates that “the selection process of tion). In this sense, both thesauri and KOS require sensitive- terms and tests under the principles of “warrants” are very ness to the inclusion of viewpoints and values shared by important to the development of any KOS.” various possible stakeholders for the represented infor- Warrants are fundamental to KOS constructions and, mation. more precisely, to thesaurus constructions. This is due to Thesauri, as reality representation systems, cannot be decisions about the inclusion or exclusion of terms, which exclusive or discriminatory of the ways in which different must be rationally based to avoid the construction of an people, communities and cultures view a particular appre- instrument with randomly collected terms that does not hension about something or someone. The moment the- represent the EL of an area of knowledge. sauri fail to include perspectives on the same object, they Warrants are many, and some authors differ on quantity, set aside a community of potential users to the infor- but in the context of this research, we consider the literary, mation system, firstly because those users cannot com- philosophical and cultural warrant. The first one, literary municate with the system and secondly because they are warrant, is the most commented one when talking about not represented or included in it. the construction of DL, as Dias (2015 12, our translation) We have witnessed cultural interactions closer and highlights: “literary warrant is a necessary condition for the closer, and eventually becoming what Garcia Canclini construction of classification systems, thesauri and other (2015) names “hybrid cultures.” In our understanding, this controlled vocabularies.” Its importance stems from the hybridization is the interaction process involving different terminology stored in the specialized literature of an area. cultures, which ends up leading to cultural merge. This per- Thus, this warrant seeks information sources to validate ception of hybrid cultures is necessary for information or- decision-making and collect terms and concepts that will ganization when we consider a diversity of people, from be part of a KOS. This is a positivist view of a term rep- different cultures, accessing, interacting and retrieving in- resentativeness for an area of knowledge. However, it is formation in the same information system. highlighted by the technical standard of the American Na- As for knowledge organization, these principles are ap- tional Standards Institute/National Information Standards propriate to the activities, but not to all the processes per- Organization (ANSI/NISO, Z39.19 of 2005) as an indic- meating this work today. These processes demand refor- ative element to ensure semantic quality. We can verify this mulation to adapt to the inclusion of cultural diversity. fact in the very concept presented for literary warrant, One possibility is to include warrants other than literary which is the “Justification for the representation of a con- and use ones, the most widely used, to include cultural war- cept in an indexing language or for the selection of a pre- rant. ferred term because of its frequent occurrence in the liter- ature.” (National Information Standards Organization 3.0 Warrants in thesaurus construction 2005, 6). The whole process involved in literary warrant seeks to find the basis for proposing a KOS in the docu- Warrants are studied in the context of KOSs, and the spe- mented titles. As KOSs are often designed for specific ar- cific literature on IS offers large scientific production re- eas, the language the expert community uses to point out lated to classification systems. However, the application of the most appropriate is sought. these warrants is also appropriate for thesaurus study and Philosophical warrant, also found as scientific warrant construction. Tennis (2005) highlights that warrants are is “consistent with scientific and philosophical consensus the rationalization of the choice of a term or concept to and is based on the authority of the academia and re- be included or not in a controlled vocabulary, as it will give search” (Guedes and Moura 2016, 83, our translation). To the necessary limits for inclusion or exclusion based on this end, KOS construction should turn to science, or to terminology. Thus, Feinberg (2010, 492) states that a war- scientific practices, for decision making regarding the cre- rant “defines the potential sources and rationale by which ation of the system. This is one way of considering and a classification designer determines the contentof the clas- building a KOS, based on greater stability and greater sification.” standardization. Thesauri are constructed and formed based on decisions With this, it is possible to have a warrant which is close that cannot be merely taken by someone or by a group’s to DL’s goals, as they aim at language standardization within wish or personality. Solid foundations for making these de- an information system. However, it is noteworthy that DLs cisions are built, which are generally related to the inclusion, seek standardization but not a regularity of language use, exclusion or relocation of terms within the thesaurus. War- since many authors comment on the need for revision and rants, therefore, “are important to validate the concepts and adequacy of DL terms. Bliss (apud Guedes and Moura 2016, terms used to represent a given domain” (Dias 2015, 10, our 83, our translation) comments that “the importance of the 642 Knowl. Org. 46(2019)No.8 P. Gomes and M. Guiomar da Cunha Frota. Knowledge Organization from a Social Perspective relationship of principles based on classical structures of Cultural warrants influence the underlying operational thought for the development of a classification system sug- rationale upon which classification systems depend gests joining practical, logical and philosophical principles in for meaningfulness and utility. To investigate cultural guiding the development of classification schemes.” warrant beyond the intuitive or observational level, Beghtol (1986) indicated that philosophical warrant was the techniques and findings of such fields as sociol- pointed out as the most suitable for KOS construction by ogy, the sociology of knowledge and social/cultural seeking in science, or in academia, the basis for identifying anthropology would have to be applied to the study what has relevance to the system. This is due to the con- of bibliographic classification systems. stancy science has on the subjects and themes. Its stability in terms of language would vary less frequently than that We understand that, to be able to include this warrant, ad- used by an area expert or users. dressing sociological techniques that cope with the cultural Finally, cultural warrant gained prominence especially part to be inserted into the KOS would be necessary. Some with Beghtol’s 1986 article “Semantic Validity: Concepts of to be indicated are those intended for data collection, in- Warrant in Bibliographic Classification Systems.” This war- terview, focus group, observation and could be used to un- rant means that “any kind of knowledge organization derstand the process of signification of the elements to be and/or representation system may be most convenient and included in the system, to make it more diverse in terms useful for people in a culture only if it is based on the as- of worldview. sumptions, values, and predispositions of the same culture” (Guedes and Moura 2016, 13, our translation). The “per- 4.0 Methodological procedures spective of cultural warrant is a way of reaching notions and ideas which are difficult to recognize by other semantic pa- The methodological strategies were systematized to allow rameters in which both user communities can identify and the analysis of thesaurus construction procedures and to KOS are able to represent abstractions of these ideas” (14). allow indications that make it more open and receptive to Cultural warrant is, therefore, the means of inserting users’ cultural diversity. Initially, we sought thesaurus construc- values into a KOS, which will afterwards be accessed by the tion models that could support the analysis. Among the users themselves. This process brings the users closer to the several found, the Integrated Methodological Model for information system, as it reflects values and predispositions Thesaurus Construction (IMMTC), in annex I, was the on some of their assumptions. most appropriate in the context of this empirical research. Pinho (2006, 64), in line with Beghtol, argues that cultural This model was proposed by Cervantes (2009) in her doc- warrant is a “way to add flexibility to knowledge organiza- toral thesis in IS. The purpose of her study was to system- tion and representation systems to encompass aspects re- atize the various models, norms and authors to create a lated to cultural diversity and make them represented.” This model integrated in steps and procedural instructions. The state is important as each group creates and is part of a spe- model structured and compiled the construction steps in cific culture. If KOS can aggregate them, these cultural one single instrument and thus facilitated and reduced re- specificities become, as the author puts it, more flexible as search efforts. However, each step proposed in the IM- they give users a perspective that is known and shared by MTC was decomposed to understand each of the indi- them. cated instructions in a more complex way. With this, there In Gracioso's (2010) understanding, the inclusion of cul- was a return to the sources used by the researcher. tural warrant would be a way to reinforce semantic relations Next, the critical reading of the model and the theoret- within KOSs. Thus, at the end of a KOS creation, there ical construct on cultural warrant was performed for a would be a relation of meanings closer to the users’ comparison and association that suited the guidelines knowledge. It is somewhat predictable that with these cul- given by construction models based on philosophical and turally closer relationships with users, information profes- literary warrant. Finally, the collected guidelines were cate- sionals could better understand the relationships of mean- gorized to facilitate the inclusion of proposals that would ing at the time of their work with information. Another allow the model to be open to cultural warrant. point Beghtol (2002) makes, which is recognized by Guedes (2016), is the understanding of cultural warrant as an um- 5.0 Analysis and discussion of results brella concept that would hold all other warrants. They are a form of complement supporting cultural warrant. This After the initial analysis of the model, four founding ele- umbrella metaphor was given Beghtol as early as 1986 and ments were identified and categorized in the process to fa- has come back with Guedes (2016) and Guedes and Moura vor understanding the data obtained from the reading of (2016). the model and its guidelines. The perceived categories Finally, Beghtol (1986, 121) indicates that: were: a) people; b) materials; c) methods; and, d) processes. Knowl. Org. 46(2019)No.8 643 P. Gomes and M. Guiomar da Cunha Frota. Knowledge Organization from a Social Perspective
This step was followed by the inclusion of elements the role of this group, their main role was to advise about that allowed the thesaurus to have cultural hospitality. With the composition of the bibliographic materials to be con- this, in each of the categories, strategies were conceived to sulted for terminological research and for guidance on the allow the inclusion of more than one voice in the final the- conceptual structure of the area. This matters as the saurus, that is, more voices in addition to science and aca- worldviews of the first group must be aligned to those be- demia perspectives, which were granted by the philosoph- longing to the second group. These understandings should ical warrant as the holders of EL of an area. be based on the pursuit of diversity and tolerance. The intention was to open the possibility of having the To bring diversity to the composition of this second participation of other social actors that contribute, in the group, the inclusion of people outside academia and sci- daily dynamics, to the construction of the EL of a certain ence is also suggested. It does not mean excluding the par- area. Limiting science and academia as legitimate holders ticipation of this collaborator profile, but the inclusion of of this language would reduce the importance of other ac- other people, such as from social movements, from the tors and open gaps that could lead to social oppression and technical practice of the area of knowledge, people who to the perpetuation of discourses that are already settled, work in institutions related to thesaurus and others. but in need to be reviewed. It is based on these four ele- With a more diverse formation of the consultant team, ments that the analysis of the obtained data was made. the exchange of knowledge, terminology, perspectives and points of view is possible, which is fundamental for de- 5.1 People signing the thesaurus hospitable to cultural diversity. The multiplicity of voices in the construction of the instru- By analyzing IMMTC, two groups of people involved in the ment does not imply the inclusion of several terms for the construction of thesaurus are verified. The first group con- same concept, or several concepts for a term, but the pos- sists of information professionals who apply the model and sibility of a diverse team to collaboratively reach consensus manage its execution. The second group are “consultants” and assist in decision making. to assure the semantic quality to the process, required by an The choices regarding the formation of the teams are instrument that conceptually represents an area of relevant to how receptive they are to differences and diver- knowledge. sity. Depending on the management of the teams, only sed- The two groups were divided and presented in separate imentary, oppressive and prejudiced forms may be con- groups, as each group has different relations with the con- ducted; however, on the other hand, it may also bring open- struction process and is not to be confused. The formation ness to cultural diversity, characteristic of today;s society. of the first group is independent of the second one, but the opposite is not true, as the second group depends on the 5.2 Materials choices made by the first one. Differentiating people into groups was important to make specific indications for each In the traditional models of thesaurus construction, the one as they are not general guidelines and would cover all terminology of a certain area is recorded on some support: involved. the materials. More specifically, through the guidelines of The group of professionals involved in the construction the analyzed model and the others that supported its elab- of the thesaurus used guidelines aimed at the formation of oration, those are bibliographic materials. a multidisciplinary team that has the viewpoint and critical At this point, the guidelines given by philosophical war- positioning directed to the worldview the thesaurus in con- rant are reviewed, but with the addition of literary warrant. struction seeks to hold. The issue involved at this moment The combination of these two warrants provides the guid- is the ethical posture of such professionals in the pursuit of ance that the materials to be used must be those created greater cultural inclusion. Knowing that the work to be done within the academia or by science, philosophical warrant, reflects on how the information system user community will and that the support and informational format would be identify. the bibliographic sources (books, specialized dictionaries, As for the second group, formed by consultants, the manuals, journals, congress proceedings and others), due guidance given by the author of the model based on the in- to the literary warrant. struments she uses is that experts be sought. There is a gap In the guidelines of the bibliographic materials produced on the understanding of who these experts are, but, assum- by science and academia, several of the instructions present ing that the basis of the instruments the author used focuses in thesaurus construction models are highlighted. It is un- on philosophical warrant, we understand that the indication derstood that the EL of an area would emerge and be cre- of the expert is one with an academic background. ated by the specialists, again, those in academia, who regu- It is noteworthy that this group is of great importance larly search in these materials for specialized terminology. on the final result of the thesaurus. In the guidelines about 644 Knowl. Org. 46(2019)No.8 P. Gomes and M. Guiomar da Cunha Frota. Knowledge Organization from a Social Perspective
However, this line of thought, especially in the human information retrieval and dissemination. One of these and social sciences, presents certain complications as it is products is the representative terms of concepts present excluded from the complexity of an area, the social actors in the consulted documents. This implies that these col- and their production of information and knowledge, lected terms will be used for the conceptual construction which contribute to the constitution, organization and of the chosen area for the thesaurus. Several authors es- maintenance of discussions that may not have been cre- tablish the need for understanding the structure of the text ated by those with an academic background, who hold the as the objective for a good reading, therefore, knowing the title of specialist but who maintain communication lan- macrostructures of the text. This implies using documents guage that is not so distinct from so many others. This with standardized structures that allow this reading with means there is specialized communication outside science predefined directions. To provide guidelines for documen- and academia, which is often set aside for the creation of tary analysis, NBR 12.676/1992 was created. It also in- thesaurus and other DLs. cludes reading the works as a way to identify relevant terms As a way of breaking this maintenance of science as an and concepts. However, a relationship of standardization EL holder of an area in the thesaurus construction process, of procedures is perceived. the inclusion of diverse sources is indicated. Possible exam- The new proposal for thesaurus construction methods is ples are to understand and collect the terminology of non- to use more open procedures. Considering Beghtol’s (1986) academic books, newspapers and magazines, social move- proposal to use research techniques from sociology, sociol- ment bulletins, interviews, documentaries, testimonials, etc. ogy of knowledge and social anthropology, the author indi- It is also necessary to think about the disruption of using cated the use of techniques used to collect data from re- written materials. Audiovisual sources can be important search in these areas to achieve an approach and an open- sources for collecting the terminology of certain areas, as ness to cultural warrant. We understand, therefore, that to the case of the Transitional Justice in Brazil, a branch that include this warrant, addressing sociological techniques that has collected testimonies of victims of the Brazilian military cope with the cultural part to be inserted into the KOS is dictatorship and which incorporates the specific worldviews necessary. Some techniques to be indicated are: interviews, and terminology in its reports but to some extent are also focus group, observation, which could be used to under- shared by science. stand the process of signification of the elements to be in- The need to reaffirm that the guidelines do not exclude cluded in the system, to make it more culturally diverse. or invalidate the importance of science and academia for the These data collection instruments would be indicated construction of thesauri is again noteworthy, but we high- for collecting the opinion of consultants, those who have light the inclusion of new actors and materials created by knowledge, not only scientific and academic knowledge, them in their movements, in their struggles and in practices but also of the analyzed area. It is a way of bringing these that involve the development of an area of knowledge. consultants closer to the thesaurus construction team. As they include data collection techniques, in this case, espe- 5.3 Methods cially opinions, the indication of techniques to understand what was collected is necessary. Part of the collected work Methods comprise the systematization for thesaurus con- is easy to understand, but the other part, which is not so struction. It is through this set of guidelines, which also explicit, needs methods that help the team understand represent the team’s choices, that will determine how close what the consultants said. The use of discourse analysis or far the final instrument will be to a quality representa- and content analysis can be effective at this phase. tion of the designed area. The main methods indicated in Following the indications of documents to be used for the IMMTC and the sources are documentary analysis and terminological collection, the next step is to analyze these technical reading of the works. The second one is part of documents to verify that their understanding of the world the procedure indicated in the first one, that is, the tech- aligns with the view the thesaurus will reflect. Again, the nical reading is a part of the procedures that comprise use of discourse analysis and content analysis techniques what the documentary analysis is. To find the most rele- are important. As books, articles, videos and other docu- vant concepts and terms within the documentary corpus, ments culturally reflect worldviews of those who have choosing a technique that allows their identification with- written them, the works need to be aligned with the instru- out having to read all the documents in their entirety is ment to be built. As going through the works to collect necessary. Thus, documentary analysis emerges, which is fundamental terms and concepts is necessary, reading con- widely used by information professionals and studied by tinues as a procedure. However, the use of a less standard- information science. ized reading is indicated, as it takes into consideration Documentary analysis process begins with reading the more diverse materials than those proposed in traditional document until it is reduced to “products” that facilitate models. Knowl. Org. 46(2019)No.8 645 P. Gomes and M. Guiomar da Cunha Frota. Knowledge Organization from a Social Perspective
Therefore, the creation of reading strategies for each understand and argue that there are knowledge areas and type of terminological source is needed. The bibliographic domains that do not respond to standardized ways in their strategies are the most common that have systematizations conceptual and terminological construction. This may not in the information science literature. However, the most di- be the case in areas such as exact sciences and biological verse sources, such as testimonials, documentaries, personal sciences, which tend to have a greater constancy of con- documents and administrative documents, need more open ceptual and terminological standards. reading strategies, as they fail to follow standardization. Extending this model to areas such as the humanities Thus, the strategies would be predefined, but not closed, and social sciences is necessary to the point that they do paths the professional would take to identify relevant terms not respond so clearly in their conceptual and terminolog- and concepts for insertion into the thesaurus. ical construction. Various actors in the field, not only aca- demia and science are involved. Social movements, institu- 5.4 Processes tions and professionals often constitute new terminologi- cal forms that end up affecting the understanding of in- The processes in the analyzed model seem to refer to the clusion and exclusion. most standardized forms possible. This must have been a In times of revision of concepts and terms that may characteristic that allowed great expansion of thesaurus represent oppressive forms of power and at the same time construction, because when the complexity of construction generate social exclusion, it is necessary to think of instru- of these instruments is restricted to standardized proce- ments that include forms of incorporating people and ter- dures, they facilitate the construction process. Standardized minological materials in their procedures that allow the processes assist in the creation of automated processes. representation of an area in an inclusive manner and with For a shift toward cultural warrant, the process needs to a different cultural perspective. consider the specificities of each area. Some areas have char- For this, thinking beyond the simplistic view of thesauri acteristics that prevent using models as standardized as IM- being instruments of semantic reduction within the con- MTC. Areas still under construction that are intertwined text of information systems is needed. Information pro- with struggles and social disputes in the material and sym- fessionals must be aware of the role this reduction plays bolic fields are a good example, as the terminology of the within the collection, and for their user community they area is difficult to understand. Disputes involve the process can seek greater diversity and inclusion of concepts and in complex decisions that in some ways represent inclusions terms. This implies making thesauri more democratic in- and exclusions. Therefore, it is always necessary to turn to struments of social visibility. forms of inclusion and to an ethical perspective guided by This research does not yet represent a conceptual diverse worldviews that fit the thesaurus’s goal. breakthrough of traditional forms of thesaurus construc- The process, with all these transformations, require pro- tion, but in its essence, it raises a need for a review of the- fessionals involved in construction who have the knowledge saurus models, rules, norms and guidelines. All this to pre- and skills to work with the indicated terminology sources, as sent more dynamic and open guidance about the under- well as the methods of collecting opinions, reading literature standing that not all areas and domains respond equally to and other works. This is all based on a critical view that al- conceptual and terminological construction. lows constant questioning about the role the thesaurus will play not only in the information organization but also how References it will identify with the user community. Barité, Mario. 2011. La garantía cultural como justificación 6.0 Conclusion en sistemas de organización del conocimiento: aproxi- mación crítica. Palabra clave: 1, no. 1: 2-11. In the course of this research we learned that cultural war- Beghtol, Clare. 1986. Semantic validity: concepts of war- rant supported the participation of diverse people, termi- rant in bibliographic classification systems. Library Re- nological materials, methods and processes for the con- sources & Technical Services 30,: 109-25. struction of thesaurus. Sources, people and materials, tra- Beghtol, Clare. 2002. A Proposed Ethical Warrant for ditionally focused on the scientific and academic point of Global Knowledge Representation and Organization view, have been expanded to the plurality of sources also Systems. Journal of Documentation 58: 507-32. recognized by information science. This leads the thesau- Boccato, Vera Regina Casari and Ricardo Biscalchin. 2014. rus to the possibility of living a diversity of points of view As dimensões culturais no contexto da construção de vo- that have been neglected in traditional models. cabulários controlados multilíngues. Revista interamericana The objective of the research was not to disqualify or de bibliotecología 37: 237-50. criticize the model proposed by Cervantes (2009) but to 646 Knowl. Org. 46(2019)No.8 P. Gomes and M. Guiomar da Cunha Frota. Knowledge Organization from a Social Perspective
Carlan, Eliana. 2010. “Sistemas de organização do conheci- Guedes, Roger de Miranda. 2016. “O princípio da garantia mento: uma reflexão no contexto da ciência da in- semântica e os estudos da linguagem.” 2016. PhD diss., formação.” Master’s thesis, Universidade de Brasília. Universidade Federal de Minas Gerais. https://bit.ly/ https://bit.ly/31WH21q 2BQrNMw Cervantes, Brígida Maria Nogueira. 2009. “A construção Guedes, Roger de Miranda and Maria Aparecida Moura. de tesauros com a integração de procedimentos termi- 2016. O princípio da garantia semântica e os estudos da nográficos.” PhD diss., Universidade Estadual Paulista. linguagem. Tendências da pesquisa brasileira em ciência da in- https://bit.ly/340yZlo formação 9, no. 2: 1-21. Cintra, Ana Maria Marques. et al. 2002. Para entender as lin- Guimarães, José Augusto Chaves and Fabio Assis Pinho. guagens documentárias. Coleção Palavra-chave. São Paulo: 2007. Desafios da representação do conhecimento: Polis. abordagem ética. Informação & informação 12: 1-21. Dias, Célia da Consolação. 2015. A análise de domínio, as National Information Standards Organization. 2005. comunidades discursivas, a garantia da literatura e outras Guidelines for the Construction, Format, and Management of garantias. Informação & sociedade: Estudos 25, no. 2: 7-17. Monolingual Controlled Vocabularies. ANSI/NISO Z39.19- Dodebei, Vera Lúcia Doyle. 2002. Tesauro: Linguagem de rep- 2005. Bethesda: NISO Press. resentação da memória documentária. Niterói: Intertexto. Pinho, Fabio Assis. 2006. “Aspectos éticos em repre- Feinberg, Melanie. 2010. "Two Kinds of Evidence: How sentação do conhecimento: em busca do diálogo entre Information Systems form Rhetorical Arguments." Antonio García Gutiérrez, Michèle Hudon e Clare Journal of Documentation 66: 491-512. Beghtol.” Master’s thesis, Universidade Estadual Paulista. García Canclini, Néstor. 2015. Culturas híbridas: Estratégias https://bit.ly/343J6pE para entrar e sair da modernidade. 4th ed. São Paulo: Edusp. Tennis, Joseph T. 2005. Experientialist Epistemology and Gracioso, Luciana de Souza. 2010. “Parâmetros teóricos Classification Theory. Knowledge Organization 32: 79-92. para elaboração de instrumentos pragmáticos de repre- Trivelato, Rosana Matos da Silva and Maria Aparecida sentação e organização da informação na web: consid- Moura. 2017. “A diversidade cultural e os sistemas de rep- erações preliminares sobre uma possível proposta resentação da informação.” In: Memória, tecnologia e cultura metodológica.” Revista de Ciência da Informação e Documen- na organização do conhecimento, ed. Fabio Assis Pinho and tação 1, no. 1: 138-58. José Augusto Chaves Guimarães. Recife: Ed. UFPE.
Appendix I: Systematization of thesaurus construction steps
INTEGRATED METHODOLOGICAL MODEL FOR THESAURUS CONSTRUCTION
Systematization of thesaurus construction steps (standardization, literature and thesaurus) – Terminographic procedures 1. Preliminary work − Choice of the thesaurus domain and language; (General guidance /Use of automated − Subdomain delimitation; equipment of data processing) − Establishment of the limits of the thematic terminologic research; − Consult with domain/subdomain expert. 2. Compilation Method − Collection of the terminologic corpus; (Compilation Approach) − Establishment of domain tree; − Expansion of representation of chosen domain. 3. Term record − Term collection and classification. 4. Term Verification − verification, classification and confirmation of terms; (Admission and exclusion of terms / − creation of definitions; Specificity) − use of specialized vocabulary to establish relationships among descriptors and re- lationships among descriptors and non-descriptors. − Organization of relations among descriptors. 5. Thesaurus presentation forms − Thesaurus presentation works. Source: Cervantes (2009, 163, our translation).
Knowl. Org. 46(2019)No.8 647 M. Barité. Towards a General Conception of Warrants: First Notes
Towards a General Conception of Warrants: First Notes* Mario Barité Universidad of the Republic, Uruguay, San Salvador 1944, CP 11200, Montevideo, Uruguay
Mario Barité is Full Professor at the Faculty of Information and Comunication, University of the Republic, Uruguay. He holds a doctorate and has a master’s degree in scientific information from the University of Gra- nada, Spain. His main research interests include knowledge organization, terminology and e-governance. He is a level I researcher of the Uruguayan National System of Researchers. Among another contributions, he has pub- lished a Dictionary of Knowledge Organization. He is editor-in-chief of Informatio Journal.
Barité, Mario. 2019. “Towards a General Conception of Warrants: First Notes.” Knowledge Organization 46(8): 647- 655. 49 references. DOI:10.5771/0943-7444-2019-8-647.
Abstract: The areas of knowledge are organized around the identification of their terms of reference and the relationships established between them. This is the rational basis of -among others- the methodology for the development of knowledge organization systems. The authority from which to select, evaluate or revise the terminology of these systems is established in relation to any of the twenty-one warrants (literary, cultural, etc.) that have been proposed and studied unequally and autonomously in the literature of the area. This paper intends to introduce initial notes and comments to advance towards an overall conception of the warrant notion. For this purpose, the expression “warrant” is studied as a word of the general language as well as a term of specialized languages. Then, the scope of application of the warrants is established. Next, each warrant is placed in one of the approaches proposed by Hjørland to categorize theories and methods (empiricism, rationalism, historicism and pragmatism). From the above, some lines of research problems are identified. A typological table that includes data on all the warrants established until now is proposed, and the first conclusions are drawn.
Received: 30 September 2019; Revised: 25 November 2019; Accepted: 25 November 2019
Keywords: warrant, warrants, knowledge organization, terminology, classification
* Presented at ISKO-Brazil 2019: Organização do conhecimento responsável: Promovendo sociedades democráticas e inclusivas (Respon- sible Knowledge Organization: Fostering Democratic and Inclusive Societies), at Umarizal (Belém), Pará, Brazil, September 2-3, 2019.
1.0 Introduction According to a definition given by Beghtol (1986, 110), which can already be considered canonical: Terms constitute the basis of the conceptual structures constructed in each discipline or subject field to present in warrant of a classification system can be thought of an organized way its main concepts, as well as its premises, as the authority a classificationist invokes first to jus- principles, theories, categories of analysis, agreements and tify and subsequently to verify decisions about what divergences (Cabré 1993; Burke 2002, 111-152). The areas classes/concepts to include in the system, in what of knowledge are organized, reviewed and updated not order classes/concepts should appear in the sched- only based on the identification of their terms of reference ules, what unit classes/concepts are divided into, but also on the relationships established between them. how far subdivision should proceed, how much and This is the rational basis of the methodology of vocabu- where synthesis is available, whether citation orders lary control operations and, therefore, of the procedures are static or variable and similar questions. for developing knowledge organization systems, such as thesauri, taxonomies and classification systems (Iyer 2012). The abovementioned is the first definition of warrant rec- The warrants constitute the invisible support, the dis- ognized in the literature of the area, so it is possible to speak guised criteria for the selection of terms for inclusion in of a late identification of the concept, especially if it is con- knowledge organization systems and information systems sidered that Hulme had already established in 1911 the initial in general with the main objective of favoring subject re- explanation of what he called “literary warrant,” one of the trieval. specific varieties of warrants. It turns out then that the spe- 648 Knowl. Org. 46(2019)No.8 M. Barité. Towards a General Conception of Warrants: First Notes cific concept was coined seventy-five years earlier than the 2.0 Analysis of the notion of warrant corresponding generic concept. Hulme mentions that liter- ary warrant is a test of validity of a heading, as an indirect 2.1 Warrant as a word of the general language and expression of what he understood as an actual warrant as a term of specialized languages (Hulme 1911, 447). Nowadays, as will be seen, there is a rec- ord of more than twenty warrants that have been proposed A first issue to be noted from the semantic point of view as support or authority in the selection of the terminology is that “warrant” is a word of the general language as well contained in knowledge organization systems, which speaks, as a term registered in the language of different specialized on the one hand, of the recognition of the validity of this fields. Thus, it appears in the terminology of domains such concept and its usefulness in the processes of construction, as banking, finance, business, constitutional, civil and com- revision and evaluation of knowledge organization systems; mercial law. The term “warrant” has also been used in ac- and on the other, the need to return to the general notion ademic texts as the justification for an argument, the evi- of warrant in order to reflect on its essence, scope and fu- dence that supports an idea (for example, Nunns, Peace ture applications. and Witten 2015) or the philosophical discussion of beliefs Although some authors after Beghtol have made pro- (Plantinga 1993). gress on the concept of warrant, especially in recent years It is not the object of this work to determine whether (Cochrane 1993; Duff 1998; Campbell 2008; Barité 2011, the word went from the general language to the specialized 2018; Bullard 2017), there is a need to constitute a more language or vice versa, although it can be presumed that homogeneous, stable and profound body of ideas. This the need to strengthen commitments, debts and obliga- work aims to be a first approximation to this objective. To tions between people, and later between people and/or in- that end, the sequence will be as follows: stitutions or states, has accompanied the parallel evolution of the word in the general language, and the term in the i) the warrant as a word of the general language and as a mentioned specialized areas. In its semantic evolution, the term of the specialized languages; expansion to meanings such as “foundation,” “justification” ii) the scope of application of the warrants; and “reason” is also included. iii) the tentative placement of the warrants in one of the In the territory of knowledge organization, the warrant four epistemological approaches proposed by Hjørland strengthens in the first place the relevance and adequacy to categorize the theories and methods in knowledge or- of a term to represent a concept, so that this term can be ganization: empiricism, rationalism, historicism and used for the classification or indexing of data, documents pragmatism; or information. That said, it would seem that the issue of iv) types of warrants and presentation of a typological ta- warrants is settled term by term. However, if the warrant ble that includes all the warrants proposed to date in will constitute an intellectual criterion to select the termi- the literature; nology of a knowledge organization system (Huvila 2006), v) the identification of research problems on warrants; which aims to obtain a consistent and updated subject rep- vi) the first conclusions on the way to a general concep- resentation of a domain, then a comprehensive view tion of the warrants in knowledge organization. should prevail. Bullard points out (2017, 76) that the
The mere statement of such an ambitious goal as the de- warrant is an element of all classification design, re- velopment of a general conception may seem excessive or gardless of whether it is named as such and regard- disproportionate. However, there are enough propositions, less of the particular technological basis of the sys- ideas and contributions of a theoretical and methodologi- tem. Indeed, warrant is a common thread across a cal nature scattered here and there in the literature of wide variety of systems ranging from traditional li- knowledge organization to begin to assemble the pieces of brary classification to in-application menus and cat- the puzzle. These brief initial notes, of a generic but com- egories for web-based collections [because] all de- prehensive nature, constitute the first result of exploratory signers of textual organizing schemas must look to research to be continued. some source of terminology.
From this broad perspective, the warrant should be seen as a tool not only for conventional resources such as a the- saurus, but also for the choice of expressions intended for any grouping of documents or data on the internet. Going a little further, their use and potential could be extrapo- lated to other areas, such as the selection of the terms to Knowl. Org. 46(2019)No.8 649 M. Barité. Towards a General Conception of Warrants: First Notes be included in a specialized dictionary, based on the same And the sixth meaning establishes that authority is every arguments and methods. “text, expression or set of expressions of a book or writing, In order to better justify the need for an integrated epis- which are cited or alleged in support of what is said” (Real temological perspective, Olson (2004) tracked the bottom Academia Española 2014, 246). Although the two meanings line of the notion of warrant in Francis Bacon’s seem to fit better with the traditional vision of warrants, knowledge classification work and its reflections on library considering the social respect for the word of scientists and classification systems. In that work, she points out that Ba- thinkers, and the veneration for the written word, the truth con created a knowledge structure based on a common is that they also allow new authority figures to fit into them: episteme on which the classificationists have been assem- the leaders of certain cultures or minority groups, the alter- bling their system designs from the nineteenth century on- native terminologies proposed by thinkers or social move- wards. And she concludes (Olson 2004, 4) that: “perhaps ments—especially the countercultural ones—or even by the we should also follow his epistemological warrant and let suppliers of new technologies. The doors can only be our classifications not only reflect knowledge, but also opened to new authority figures if new warrants different have a role in directing the creation of new knowledge.” from the traditional ones are introduced, warrants which This neo-positivist perception may have led, in a more or meet the needs of subject representation of groups, cultures less intuitive way, the classificationists to fuel—over a cen- or subcultures of low visibility. tury and a half—an allegedly common epistemological vi- sion, in order to validate the idea that the precepts of sci- 2.2 Scope of application of the warrants ence are objective, neutral and universal in scope. In a more hidden way perhaps, it was gradually estab- Beghtol’s groundbreaking definition (1986), besides being lished that the justification of the terminology of a extremely detailed, is very accurate. For example, it is a merit knowledge organization system should be taken in full, ei- of hers to propose the incidence of the warrant at two levels ther from the formal language of science and science edu- or times: that of the initial justification (that is, when creat- cation (Bliss 1929) or from the terminological evidence pro- ing or selecting a descriptor or subject heading) and in the vided by the documentation (Hulme 1911). This traditional verification (for example, at the time of the evaluation of scheme gradually broke down with the proposal of new the quality and relevance of a term). This implies that the warrants (cultural, user, organizational, among others), warrant can be used as guidance and as a tool in the three which constituted breakpoints in two ways: 1) in relation to most important processes that a knowledge organization the alleged objectivity and neutrality of the language of sci- system can go through: its construction, its evaluation and ence; and, 2) regarding the desirability of maintaining a sin- its revision. Therefore, the tuning of specific methodologi- gle warrant for all the terminology selected in the process of cal devices to choose, assess and replace or update their ter- developing a knowledge organization system. minology fits here. When Beghtol states that the warrant is the authority that Another of the high points of Beghtol’s work is the state- a classificationist invokes (Beghtol 1986, 110), she places in ment—never contested in the literature of the area—that that authority the source of legitimacy and the ultimate basis the warrant must be applied at all stages of the design of a for decision-making regarding the terminology to be in- knowledge organization system, namely: cluded in or excluded from a knowledge organization sys- tem. She says it unequivocally: the warrant is the authority. i) in the selection of the terms of classification and in- This principle of authority may be firmly established in the dexing; terminology chosen by those responsible for a knowledge ii) in the selection of the relationships established be- organization system, or it may be more or less blurred, as a tween them, an issue also mentioned later by other au- consequence of the selection of ambiguous, generic or hy- thors (Rowley 1987; Barité et al 2015, 77); brid terms, or not sufficiently representative, according to iii) in the arrangement of the terms of the facets, thus the greater or lesser degree of methodological rigor in the resuming the original application of the literary war- selection and systematization of the terminology. A more or rant proposed by Ranganathan in his Prolegomena less corresponding relationship should then be presumed (Ranganathan 1967, 196); between the strength and the adequacy of the authority or iv) in the choice of criteria for the subdivision of matters; warrant and the quality of the final terminology product. v) in the determination of the specificity levels; Two of the six definitions established by the Dictionary vi) in the application of synthesis mechanisms (as in the of the Real Academia Española for the entry “authority” are choice of auxiliary tables or the signs of combination applicable in this context. The third meaning says “prestige of issues), the selection of syntax devices; and and credit attributed to a person and/or institution for its vii) in the citation order of matters (Beghtol 1986, 110) legitimacy or its quality and competence in some matter.” 650 Knowl. Org. 46(2019)No.8 M. Barité. Towards a General Conception of Warrants: First Notes
At this point, it should be noted that warrants can become ble way to the more general foundations of the sciences epistemological references, organizational criteria or tools and disciplines. Besides, these intellectual tools can be ap- at the service of a terminology selection method. plied to any topic of interest as an object of study for It is noted that the type of warrant applied to the selec- knowledge organization (as is the case with warrants). tion of classification or indexing terms may affect not only The four epistemological-based approaches proposed the decision to include or exclude terms but also the dis- by Hjørland are: 1) the empiricism, which is justified by the tinction between authorized and unauthorized terms. In- data coming from the set of observations and their corre- deed, the decision that is taken, for example, to establish sponding inductions; 2) the rationalism, which proposes in a thesaurus that the term “small states” will be a de- the development of knowledge based on principles of scriptor, while its synonym “microstates” will be consid- pure logic or pure reason and which relies substantially on ered a non-descriptor, must be based on a criterion sup- deductive processes; 3) the historicism, which promotes ported by a warrant. The way to warrant relationships be- the organization of knowledge based on chronological, tween terms has been an issue just outlined in the literature, evolutionary and/or contextual studies of each field of both from a theoretical and methodological point of view, knowledge; and finally, 4) the pragmatism, in whose es- so it constitutes an open line for research. In these notes, sence the analysis of reality is based on the determination as a first statement subject to review, it is suggested that of values, goals and consequences (Hjørland 2003, 2006, the warrant procedure for hierarchical relationships be- 2013). In this exploratory work—in the table presented in tween terms should be different from the procedure for Appendix A—each warrant will be related to at least one associative relationships. of the four approaches mentioned on account of future Indeed, the hierarchical relationships established in a particular analyses. knowledge organization system are usually those formally stated and accepted in different disciplines. They are stable, 2.4. Warrant types proven and not casual relationships. On the other hand, the systems usually provide the possibility for the classifier, In a recent review of the existing literature on the topic, a the indexer or the end user to establish an associative rela- total of twenty-one warrants proposed, named and used tionship between two issues, perhaps because it may be in knowledge organization and close subject fields (Barité temporary or provisional. Of course, this division of qual- 2018, 528) were registered. In that work, there is a synthe- ities between hierarchical and associative relationships is sis table with the relation of all the warrants. Not all the not absolute, at least as regards a large set of associative authors explicitly coined the term “warrant” although it relationships that are firmly established in the world of was clear in the respective texts that, in essence, they were knowledge and documentation (for example, the relation- talking about the authority invoked to represent ships between church and state, or the relationships be- knowledge through descriptors, headings, keywords, clas- tween certain plant or animal substances and the treatment sification numbers, taxa or others symbols. For the pur- of diseases). If this provisional starting point is accepted, poses of this paper, the aforementioned table (which due the warrant for hierarchical relationships that are estab- to its length is presented—as we said—in Appendix A), lished in a knowledge organization system should more was revised and expanded, with the aim of presenting the reasonably come from the formal classifications of the dis- twenty-one warrants in a single graphic expression, clear ciplines (what is known as academic warrant). While to jus- and exhaustive, with the following data: name in English, tify an associative relationship the most appropriate war- name in Spanish, author who proposed it and the year of rant for each case should be identified. coinage. A column of comments was also added as well as another column that provisionally places each warrant in 2.3 Warrants and epistemological theories one of the four epistemological-based approaches out- lined above. In the table, the warrants are arranged in In different works, Hjørland (2003, 2006, 2013, among chronological order of proposal. others) has given evidence of the importance of associat- Since the concept of literary warrant (Hulme 1911) was ing epistemological approaches to the analysis of theories introduced, of the twenty remaining warrants, those most and methodologies in knowledge organization. The most frequently referenced are cultural warrant (Lee 1976; evident advantage of this procedure is that each theory or Beghtol 2002a), academic warrant (Bliss 1929; Sachs and set of ideas and each method can be inserted into a more Smiraglia 2004) and user warrant (Lancaster 1977; Hjør- general category of analysis, thus taking advantage of all land 2013). Basically, they are distinguished, because they the accumulated flow of reflection made from each epis- invoke different sources of authority to collect terms: the temological approach. Thus, information science (in par- language of communities with their own cultural or local ticular knowledge organization) is linked in an approacha- identity (cultural warrant), the formal vocabulary of disci- Knowl. Org. 46(2019)No.8 651 M. Barité. Towards a General Conception of Warrants: First Notes plines, consensus among specialists and expert opinion utilities that each warrant can offer are usually found. This (academic warrant) or the expressions that users use in is the case of cultural warrant, user warrant, organizational their searches (user warrant). The other warrants, however, warrant, academic warrant and more recently indigenous have only been sporadically dealt with, and there is no suf- warrant. A third group brings together the warrants that ficient theoretical, methodological and critical reflection have been proposed and/or mentioned occasionally and on them. that do not have significant subsequent development, as in Anyhow, the proposal of such a high number of war- the case of market, structural or autopoietic warrants. rants, which implies the resort to their corresponding— These last two groups of warrants do not have enough and different—sources of authority, may be due to some critical analysis yet, and they need it. (or several, or all) of the following causes: From the methodological and applicative point of view, there is a wide research scope on a series of issues that i) the insufficiency of each particular warrant to decide have no definitive answer, and in some cases, not even par- in each instance in which it is required to select termi- tial hypotheses or interpretations. These questions can be nology to represent knowledge and retrieve infor- grouped into two categories: those related to the applica- mation; tion of a warrant autonomously and those that propose ii) as a consequence of the above, it can, therefore, be the convenience of combining two or more warrants to stated that there are no universal solutions, but only support the box of terms of a knowledge organization options of scope and partial effectiveness to justify ter- system or an information system. minology; Among the first, there appear the following questions: 1) iii) in accordance with what is established by one of the How do we decide the most appropriate warrant for each premises of knowledge organization, the same articu- system?; 2) How do we guide the choice of one warrant and lated set of knowledge can be organized in n number not another one?; 3) What methodologies does each warrant of ways (Barité 2001, 48-49), according to the possibil- offer for its application?; 4) In what thematic, documentary ity of using different epistemological or practical per- or information contexts can a warrant be applied?; 5) How spectives of that set, which implies that each of these is the “performance” of a warrant evaluated on the basis of perspectives can be supported by a particular type of the principles of consistency, exhaustiveness, thematic ade- warrant; quacy, linguistic adequacy and other indicators that can be iv) the increase of the issues related to the subject repre- proposed?; 6) How do technological advances contribute to sentation of knowledge in digital information environ- or hinder the application of warrants based on algorithmic ments, which are geared towards setting, tagging and mechanisms of automatic or semi-automatic subject assign- translating the terms of any field or discipline to favor ment?; 7) What theoretical principles and methodologies re- its search and access. lated to warrants are valid in digital information environ- ments?; 8) How are warrants linked to natural language in- 2.5 Identification of research problems on warrants dexing?; and, 9) On the other hand, can it be proven that the selection of a single warrant ensures the terminological con- The status of the situation that has been outlined in this ex- sistency required for the system to be useful for users? In ploratory research regarding the works related to warrants relation to the latter, it is pertinent to wonder, given the va- in knowledge organization allows us to identify three kinds riety of warrants that have been proposed by different au- of matters that could be subject to research: theoretical, thors for over a century, if they can be combined or com- methodological and applicative issues. From a theoretical plemented to obtain a better quality and adequacy of the point of view, the exhaustive review of literature carried out terminology, or if some exclude others, and in the latter case, to support this work shows that there is a more or less con- which do and which do not and why. solidated body of accumulated knowledge on literary war- Opinions in the literature of the area are divided. rant: canonical articles (Rodríguez 1984; Beghtol 1986), Svanberg believes in the complementary use of warrants but comprehensive studies and postgraduate theses (Barité 2011, also that some may be opposed to others, without providing 2018), regular production in recent years (Howarth and Jan- further explanations (Svanberg 1996). Bullard, on the other sen 2014; Bullard 2017), as well as a recurrent analysis of the hand, expresses (2017, 77) that “the various warrants availa- application of literary warrant to the Library of Congress Clas- ble to classification designers represent contradictory posi- sification (for example, Hallows 2015) or the Dewey Decimal tions in classification theory yet they compete and are com- Classification (for example, Vizine-Goetz and Beall 2004). bined by classification designers in daily practice.” After tak- A limited number of works has been devoted to a sec- ing a position on the literary, academic, user and ethical war- ond group of warrants, in which generic introductions and rants, and discussing the possible compatibilities and incom- general approaches to the characteristics, purposes and patibilities between them, she states (2017, 77) that “inevita- 652 Knowl. Org. 46(2019)No.8 M. Barité. Towards a General Conception of Warrants: First Notes ble compromises of daily classification work” [require] “the creased on an international level. However, a reorganiza- interaction between warrants.” tion of the area is necessary. It is noted that there is a need Huvila believes that two or more warrants may be op- to build an overall view, and to go in-depth in a series of posed to each other, but he proposes to incorporate the con- issues that have been treated generically so far. It is also cept of hospitality, as reinterpreted by Beghtol (2002a), be- necessary to induce from what has been produced rules, cause (Huvila 2006, 60) it “may be used to denote an ability premises, principles and methods that may be common to to incorporate both intra and inter warrant differences i.e. all warrants or that require their proper specification. eventual changes within and between individual warrants.” Longer term works (monographs, books, postgraduate Wan-Chen Lee, on the other hand, generates ideas to under- theses) that focus on this subject are also required. stand the nature of the conflict between several semantic On the way to a general conception of the warrants, warrants, and offers some negotiation alternatives for their this exploratory work has revealed—without exhausting use and combination, within the framework of the evalua- the matter—different points of conflict, discussion or ex- tion processes of knowledge organization systems (Lee change of ideas around the warrants, and in particular, an 2017). important number of questions that could guide future re- From this brief review, there appear to be many blind search on the subject have been formulated. Perhaps the spots that can be used for research on the theory, methods time has come when instead of thinking about expanding and applications of warrants, in an area in which the emer- the list of warrants, it would be more productive to devote gence of new types of knowledge organization systems has greater conceptual, methodological and applicative con- been constant in the last twenty-five years (ontologies, web tent to each of them, since they mostly have scarce literary taxonomies, folksonomies and social classifications, among warrant to support them. others), as well as technological innovations that have had The establishment of the relationship between warrants an impact on customs and habits in relation to the search of and epistemological theories is also relevant to promote a and access to information. greater conceptual depth as well as to provide more sup- port to field studies. Perhaps the most significant aspect to 3.0 Conclusions promote research in the area is that the topic of warrants remains particularly valid due to the importance assigned Warrants are currently seen as an essential component in nowadays to subject retrieval on the internet, databases the process of construction, evaluation and revision of and databanks and other sources and information systems, knowledge organization systems, to the extent that their linked to science, commerce, e-government, culture and proper understanding and application should ensure con- entertainment industries. Their projection and utility are, sistent terminology, updated and adjusted to the purposes therefore, sufficiently consolidated. of system designers, and users’ information needs. More- over, warrants can be used as tools to guide natural lan- References guage indexing, contributing to correct the undisciplined tagging of social classifications, or they can contribute to Andersen, Jack. 2015. “Re-Describing Knowledge Organi- the selection of terms to be defined in a specialized dic- zation: A Genre and Activity-Based View.” In Genre The- tionary, among other possible uses. There are no substan- ory in Information Studies, ed. by Jack Andersen. Bingley: tial differences regarding the definition of the notion of Emerald, 13-42. warrant. The authors seem to agree that these are theoret- Barité, Mario. 2018. “Literary Warrant.” Knowledge Organiza- ical-methodological criteria that guide the selection of ter- tion 45: 517-36. minology in all information contexts where subject repre- Barité, Mario. 2011. “La garantía literaria como herramienta sentations are needed, and they are assigned as their main de revisión de sistemas de organización del conoci- task to justify the inclusion, weighting or exclusion of miento: Modelo y aplicación.” PhD diss., Universidad de terms. Granada. The root of the issues faced in this work has a promis- Barité, Mario. 2001. “Organización del conocimiento: Un ing basis: in the last twenty-five years, the academic pro- nuevo marco teórico-conceptual en biblioteconomía y duction on this topic has increased and has been greatly documentación.” In Educação, Universidade e Pesquisa: Tex- enriched, and in its diversity of approaches it has left a par- tos completos do III Simpo ́sio em Filosofia e Ciencia-Paradigmas ̂ ticularly fertile ground for the discussion of ideas. In the do Conhecimento no Final do Milenio ̂ . Marília: UNESP; mentioned period, thirteen different warrants have been FAPESP, 35-60. proposed, a regular work flow has been generated (both in Barité, Mario et al. 2015. Diccionario de organización del conoci- the form of journal articles and conference papers) and miento: Clasificación, indización, terminología. 6th rev. ed. Mon- the critical mass studying the warrants has significantly in- tevideo: CSIC. Knowl. Org. 46(2019)No.8 653 M. Barité. Towards a General Conception of Warrants: First Notes
Beghtol, Clare. 2002a. “Universal Concepts, Cultural War- Fraser, W. J. 1978. “Literary, User and Logical Warrants as rant and Cultural Hospitality.” In Challenges in Knowledge Indexing Constraints.” In The Information Age in Perspective: Representation and Organization for the 21st Century: Integration Proceedings of the ASIS Annual Meeting: 41st Annual Meeting, of Knowledge across Boundaries Proceedings of the Seventh Inter- New York, New York, November 13-17, 1978, comp. Everett national ISKO Conference, 10-13 July 2002 Granada, Spain, H. Brenner. White Plains, NY: Knowledge Industry Pub- ed. Maria J. López-Huertas and Francisco J. Munoz- lications, 130-32. Férnandez. Advances in Knowledge Organization 8. Hallows, K. M. 2015. “It’s All Enumerative: Reconsidering Würzburg: Ergon, 45-9. Library of Congress Classification in U.S. Law Librar- Beghtol, Clare. 2002b. “A proposed Ethical Warrant for ies.” Law Library Journal 106: 85-99. Global Knowledge Representation and Organization Hjørland, Birger. 2013. “Theories of Knowledge Organi- Systems.” Journal of Documentation 58: 507-32. zation-Theories of Knowledge.” Knowledge Organization Beghtol, Clare. 1986. “Semantic Validity: Concepts of 40: 169-81. Warrant in Bibliographic Classification Systems.” Li- Hjørland, Birger. 2006. “Empiricism, Rationalism and Pos- brary Resources & Technical Services 30: 109-23. itivism in Library and Information Science.” Journal of Bliss, Henry E. 1929. The Organization of Knowledge and the Documentation 61: 130-55. System of the Sciences. New York: Holt. Hjørland, Birger. 2003. “Fundamentals of Knowledge Or- Budd, John M. and Daniel Martínez Ávila. 2016. “Epistemic ganization”. In Tendencias de investigación en organización del Warrant for Categorizational Activities in Knowledge conocimiento: Trends in Knowledge Organization Research, ed. Organization.” In Knowledge Organization for a Sustainable José Antonio Frías and Críspulo Travieso. Aquilafuente World: Challenges and Perspectives for Cultural, Scientific, and 51. Salamanca: Ediciones Universidad de Salamanca, Technological Sharing in a Connected Society: Proceedings of the 83-116. Fourteenth International ISKO Conference 27-29 September Hjørland, Birger. 2017. “Subject (of Documents).” 2016 Rio de Janeiro, Brazil, ed. José Augusto Chaves Knowledge Organization 44: 55-64. Guimarães, Suellen Oliveira Milani and Vera Dodebei. Howarth, Lynne C. and Eva Hourihan Jansen. 2014. “To- Advances in Knowledge Organization 15. Würzburg: Er- wards a Typology of Warrant for 21st Century gon, 142-5. Knowledge Organization Systems.” In Knowledge Organi- Bullard, Julia 2017. “Warrant as a Means to Study Classifica- zation in the 21st Century: Between Historical Patterns and Fu- tion System Design.” Journal of Documentation 73: 75-90. ture Prospects: Proceedings of the Thirteenth International ISKO Burke, Peter. 2002. Historia social del conocimiento: De Guten- Conference 19-22 May 2014, Kraków, Poland, ed. Wiesław berg a Diderot. Barcelona: Paidós. Babik. Advances in Knowledge Organization 14. Würz- Cabré, María Teresa. 1993. La terminología: teoría, metodología, burg: Ergon, 216-21. aplicaciones. Barcelona: Antártida. Hulme, E. Wyndham. 1911. “Principles of Book Classifi- Campbell, D. Grant. 2008. “Derrida, Logocentrism and the cation: Chapter III: On the Definition of Class Head- Concept of Warrant on the Semantic Web.” In Culture ings, and the Natural Limit to the Extension of Book and Identity in Knowledge Organization: Proceedings of the Tenth Classification.” Library Association Record 13: 444-9. International ISKO Conference 5-8 August 2008 Montréal, Huvila, Isto. 2006. The Ecology of Information Work: A Case Canada, ed. Clément Arsenault and Joseph T. Tennis. Ad- Study of Bridging Archaelogical Work and Virtual Reality vances in Knowledge Organization 11. Würzburg: Er- Based Knowledge Organisation. Åbo: Åbo Akademi Univer- gon, 222-28. sity Press. Cochrane, Pauline. 1993. “Warrant for Concepts in Classifi- Iyer, Hemalata. 2012. Classificatory Structures: Concepts, Rela- cation Schemes.” In Proceedings of the 4th ASIS SIGICR tions and Representation. Würzburg: Ergon. Classification Research Workshop, ed. Philip J. Smith, Clare Lancaster, F. Wilfred. 1977. “Vocabulary Control in Infor- Beghtol, Raya Fidel and Barbara H. Kwasnik. Advances mation Retrieval Systems.” In Advances in Librarianship. in Classification Research 4. Medford, NJ: Information Vol. 7, ed. Melvin J. Voight and Michael H. Harris. New Today for the American Society for Information Science, York: Academic Press, 1-40. 57-68. Lee, Joel M. 1976. “E. Wyndham Hulme: A Reconsidera- Cutter, Charles A. 1876. Rules for a Dictionary Catalog. Wash- tion.” In The Variety of Librarianship: Essays in Honour of ington, D.C.: Government Printing Office. John Wallace Metcalfe, ed. W. Boyd Rayward. Sandy Bay, Doyle, Ann Mary. 2013. “Naming, Claiming, and (Re)creat- Tasmania: Library Association of Australia, 101-13. ing: Indigenous Knowledge Organization at the Cultural Lee, Wan Chen. 2017. “Conflicts of Semantic Warrants in Interface.” PhD diss., University of British Columbia. Cataloging Practices.” In Proceedings from North American Duff, Wendy. 1998. “Harnessing the Power of Warrant.” Symposium on Knowledge Organization. Vol. 6, 231-8. American Archivist 61: 88- 105. doi:10.7152/nasko.v6i1.15242 654 Knowl. Org. 46(2019)No.8 M. Barité. Towards a General Conception of Warrants: First Notes
Mai, Jens-Erik. 2011. “Folksonomies and the New Order: Rodríguez, Robert. D. 1984. “Hulme's Concept of Literary Authority in the Digital Disorder.” Knowledge Organiza- Warrant.” Cataloging & Classification Quarterly 5, no. 1: 17- tion 38: 114-22. 26. Martínez-Ávila, Daniel. 2012. “DDC-BISAC Switching as Rowley, Jennifer E. 1987. Organising Knowledge: An Introduc- a New Case of Reader-interest Classification.” PhD tion to Information Retrieval. Aldershot: Gower. diss., Universidad Carlos III de Madrid. Sachs, Moshe Y. and Richard P. Smiraglia. P. 2004. “From National Information Standards Organization. 2010. Guide- Encyclopedism to Domain-Based Ontology of Know- lines for the Construction, Format and Management of Monolin- ledge Management: The Evolution of the Sachs Classifi- gual Controlled Vocabularies. National Information Stand- cation (SC).” In Knowledge Organization and the Global Infor- ards Series. Bethesda, MD: NISO Press. ANSI/NISO mation Society: Proceedings of the Eighth International ISKO Z39.19-2005 (r2010). Conference, 13-16 July 2004 London, UK, ed. Ia C. McII- Nunns, H., R. Peace and K. Witten. 2015. “Evaluative Rea- waine. Advances in Knowledge Organization 9. Würz- soning in Public-Sector Evaluation in Aotearoa New burg: Ergon, 167-72. Zealand: How Are We Doing?” In Evaluation Matters: He Svanberg, M. 1996. “Classification, Warrants and Princi- Take Tō Te Aromatawai 1: 137-63. doi:10.18296/em.0007 ples.” Swedish Library Research 2, no. 3: 66-75. Olson, Hope A. and Dennis B. Ward 1998. “Charting a Svenonius, Elaine. 2000. The Intellectual Foundation of Infor- Journey Across Knowledge Domains: Feminism in the mation Organization. Cambridge MA: MIT Press. Decimal Dewey Classification.” In Structures and Rela- Tennis, Joseph T., Katherine Thornton and Andrew Filer. tions in Knowledge Organization: Proceedings of the Fifth Inter- 2012. “Some Temporal Aspects of Indexing and Clas- national ISKO Conference, Lille, France, August 25-29, 1998, sification: Toward a Metrics for Measuring Scheme ed. Widad Mustafa El-Hadi. Advances in Knowledge Change.” In Proceedings of the 2012 iConference, Toronto, Organization 6. Würzburg: Ergon, 238-44. Ontario, Canada, February 07-10, 2012. New York: ACM, Olson, Hope A. 2004. “Bacon, Warrant and Classification.” 311-6. doi:10.1145/2132176.2132216 In ASIS&T SIG-CR Workshop, November 13, 2004. Vizine-Goetz, Diane and Julianne Beall. 2004. “Using Lit- Plantinga, Alvin. 1993. Warrant and Proper Function. New erary Warrant to Define a Version of the DDC for Au- York: Oxford University Press. tomated Classification Services.” In Knowledge Organiza- Rafferty, Pauline and Rob Hidderley. 2007. “Filckr and tion and the Global Information Society: Proceedings of the Democratic Indexing: Dialogic Approaches to Index- Eighth International ISKO Conference, 13-16 July 2004 Lon- ing.” Aslib Proceedings 59: 397-410. don, UK, ed. Ia C. McIIwaine. Advances in Knowledge Ranganathan, S. R. 1967. Prolegomena to Library Classification. Organization 9. Würzburg: Ergon, 147-52. 3rd ed. Bombay: Asia Publications. Ward, M. 2000. “Phenomenological Warrant: The Case for Real Academia Española. 2014. Diccionario de la lengua espa- Working from the User’s Viewpoint.” Managing Infor- ñola. 23rd ed. Buenos Aires: RAE-Espasa. 2 vol. mation 7, no. 9: 68-71.
Appendix A: Typological table of warrants
Name in English Name in Spanish Author and year Comments Approach
Antecedent of the user Usage Uso Cutter, 1876 Empiricism warrant
Literary warrant Garantía literaria Hulme, 1911 Empiricism
Scientific/philosophical Garantía científico / Antecedent of the and educational warrant filosófica y educacional- Bliss, 1929 Rationalism academic warrant (consensus) consenso
Cultural warrant Garantía cultural Lee, 1976 Pragmatism
User warrant Garantía de usuario Lancaster, 1977 Empiricism
Logical warrant Garantía lógica Fraser, 1978 Rationalism
Garantía orientada a la Maybe a type of user Request oriented warrant Soergel, 1985, p. 230 Empiricism consulta o solicitud warrant Knowl. Org. 46(2019)No.8 655 M. Barité. Towards a General Conception of Warrants: First Notes
Name in English Name in Spanish Author and year Comments Approach
Generic name given to the literary, cultural, user Empiricism Rationalism Semantic warrant Garantía semántica Beghtol, 1986 and scientific/philosophical Pragmatism and educational warrants
Maybe a type of cultural Gender warrant Garantía de género Olson and Ward, 1998 Pragmatism warrant
Phenomenological Garantía fenomenológica Ward, 2000 Empiricism warrant
Structural warrant Garantía estructural Svenonius, 2000 Rationalism
Related to the cultural Ethical warrant Garantía ética Beghtol, 2002b Pragmatism warrant
Similar to the Academic warrant (also Garantía académica Sachs and Smiraglia, 2004 scientific/philosophical Rationalism named scholarly warrant) and educational warrant
National Information Organizational warrant Garantía organizacional Standards Organization, Empiricism 2005
Based on Rafferty and Autopoietic warrant Garantía autopoiética Mai, 2011 Hidderley, 2007. Maybe a Empiricism type of user warrant
Tennis, Thornton and Textual warrant Garantía textual Empiricism Filer, 2012
Market warrant Garantía de mercado Martínez Ávila, 2012 Empiricism
A type of cultural Indigenous warrant Garantía indigenista Doyle, 2013 Pragmatism warrant
A type of cultural Empiricism Genre warrant Garantía de géneros Andersen, 2015 warrant Pragmatism
Budd and Martínez Ávila, Rationalism Epistemic warrant Garantía epistémica 2016 Pragmatism
Policy warrant A type of cultural Rationalism (corresponding to policy Garantía en políticas Hjørland 2017, warrant ? based indexing) Pragmatism
Source: Barité (2018), revised, modified and expanded table for this paper.
656 Knowl. Org. 46(2019)No.8 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process
Changing Perspectives on Classification as a Knowledge-Representation Process † Barbara H. Kwaśnik Syracuse University – School of Information Studies, Hinds Hall, Syracuse New York 13244-1100, United States,
Barbara H. Kwaśnik taught in the areas of knowledge organization, theory of classification, and information science. Her main research interests are formal aspects of classification structure, personal information manage- ment, browsing, and how classifications intersect with everyday human endeavor. Previous research, with pro- fessor Kevin Crowston, included investigating whether genre information can help in searching. Most recently she is working with Brian Dobreski on the issues of personhood in knowledge representation. She retired in 2018 from the School of Information Studies at Syracuse University, Syracuse NY.
Kwaśnik, Barbara H. 2019. “Changing Perspectives on Classification as a Knowledge-Representation Process.” Knowledge Organization 46(8): 656-667. 37 references. DOI:10.5771/0943-7444-2019-8-656.
Abstract: No matter how immutable a classification may seem, it is, after all, an artifact of the human imagination and functions in a particular place and time. The author describes her personal inquiry into classification as a knowledge-representation process. She traces her changing perspectives on how classifications should be viewed and evaluated by posing the following questions: 1) How does the classification process enable or constrain knowing about something or discovering something we did not already know?; 2) In what ways might we develop classifications that enhance our ability to discover meaningful information in the information stores that form a part of our scholarly as well as our everyday lives?; and 3) How might classifications mask or distort knowledge, and how might they serve to disenfranchise people and ideas? These questions are considered through a discussion of classifi- cation structures, personal classification, the link of classification to theory, everyday working classifications, translation of classifications, cognitive aspects, browsing, genres, warrant, and the difficulties of navigating complex ontological commitments. The through thread is the importance of context, because classifications can only be seen with respect to the human endeavors that generate them.
Received: 13 November 2019; Accepted: 16 November 2019
Keywords: classification, classifications, information, knowledge
† In writing about my research trajectory, I was repeatedly brought up by the awareness of how many people were involved. Here’s my opportunity to thank all of them: not only the groundbreakers who were, and continue to be, my intellectual inspiration, but also my professors and dissertation committee at Rutgers University, headed by James Anderson, as well as the colleagues on a day-to-day basis, the friends who bore my annoying habit of viewing just about any issue as a classification problem. I have been fortunate that my research and teaching at the School of Information Studies at Syracuse University have been closely intertwined, which is an increasingly rare privilege. The engagement of hundreds of students who brought so much to my classes was the source of my best examples and my deepest reflections. The International Society for Knowledge Organization and the Special Interest Group for Classification Research in ASIS&T were my home where I was welcomed as a young scholar and encouraged to develop my own trajectory. The influences and opportunities for a life of the mind were constant and enveloping. My most sincere thanks, though, go to my doctoral student co-authors, who participated with me in exploring interesting questions and helping me translate what we learned into the work summarized in this essay. My work on genres with Kevin Crowston was funded by an NSF grant: Crowston, Kevin (PI). How Can Document-Genre Metadata Improve Information Access for Large Digital Collections? NSF IIS Grant 04–14482, Jan. 2005 – Jan. 2007.
1.0 Introduction they help visualize an overall view by showing clusters and areas of density and gaps. There are decisions made about We are incapable of “not” classifying. There are classifica- “first cuts,” scope, definition, and relationships among the tions everywhere, both formal and ad hoc; both enduring parts, all of this rendering a representation of some area and ephemeral. After decades of studying and teaching I of our lives and what we know about it. now understand better how classification both in and out- My formal interest in classification goes back at least side the library is not foremost about being tidy, but rather, four decades, when as a beginner cataloger I realized that about having a tool for seeing the world and understanding behind the library schemes used to organize resources it. Our efforts at creating and using classifications serve to physically on shelves was a world of intellectual richness enable more efficient communication; they show patterns; and complexity that barely tipped the surface. The ques- Knowl. Org. 46(2019)No.8 657 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process tions I asked then persist: 1) How does the classification a classification that is thrown together without any concep- process enable or constrain knowing about something or tual glue to hold it together is typically wobbly as a discovering something we did not already know?; and, 2) knowledge-representation structure. Overall, I believed that, In what ways might we develop classifications that en- indeed, a classification itself could be construed as a the- hance our ability to discover meaningful information in the ory—a structure of entities and their specified relationships information stores that form a part of our scholarly as well (Kwaśnik 1994). Whether it was good or flawed depended as our everyday lives? (Kwaśnik 1999). To that I have on the value of the conceptual scaffolding. added, over the years the thorny questions of how classi- Following on this analogy of classification to theory, I fications may mask or distort knowledge, and how they observed that the process of classification can be used in a might serve to disenfranchise people and ideas. In the fol- formative way during the preliminary stages of inquiry as a lowing sections I describe the various threads of my in- heuristic tool in discovery, analysis, and in theorizing (Davies quiry into classification as a knowledge-representation 1989). Once concepts gel and the relationships among them process and my changing perspective on the process and become understood, a classification can be used as a rich the outcomes. It is not a straight line, even though I must representation and is thus useful to communication and in present it that way. Rather, the various aspects inform each generating a fresh cycle of exploration, comparison, and other and lead to some insights, but also to fresh questions theorizing. Kaplan (1967) states that “theory is not the ag- that as of now I am not able to answer. gregate of the new laws but their connectedness, as a bridge consists of girders only in that the girders are joined to- 2.0 Knowledge and classification gether in a particular way” (297). I believed that a classifica- tion works in much the same way, connecting concepts in a I draw a distinction between merely observing, perceiving, useful structure. If successful, it is, like a theory, descriptive, or even describing things and truly knowing them. To explanatory, heuristic, fruitful, and perhaps also elegant, par- know implies a process of integration of facts about ob- simonious, and robust (Kwaśnik 1994). jects and the context in which the objects and processes This led me to believe that classification is somehow a exist (Kwaśnik 1999, 23). It is not enough to know about fundamental aspect of nature, sort of like the Fibonacci Se- the things but, rather, it is the relationship of one thing to quence, and that the “lawful” integration of theory and clas- another that creates the deeper understanding. We are fa- sification yielded the most robust schemes. There are exam- miliar with how knowledge discovery and creation in the ples to suggest this, the periodic table of the elements being sciences may follow traditional processes, such as explora- one. Here we have a classification that has endured through tion, observation, description, analysis, and synthesis, as different theoretical explanations and continues to be useful well as testing of phenomena and facts. It may also follow to this day. The scheme, which started in the nineteenth cen- more interpretive paths, which are, nevertheless, based on tury as an observation of the regular change in atomic evidence that is collected and appraised following consen- weight among elements, eventually yielded a pattern that led sual norms. All of this is conducted within the communi- to the discovery of new elements, and through the lens of cation framework of a research community with its ac- increasing theoretic understanding an explanation of why cepted methodology and set of techniques (Kwaśnik 1999, this occurred. From my perspective, it was a comforting 23). Bronowski (1978) describes how even in the empirical thought that while the principles of a classification could sciences, though, the process is not entirely rational but is evolve, the underlying concept remained solid. often sparked and then fueled by insight, hunches, and I have come to believe that this might be a dangerous leaps of faith. Moreover, research is always conducted belief in the sense that not everything sorts itself out as within a political and cultural reality (Olson 1998). beautifully as the elements in the periodic table seem to do. In my early thinking in the 1980s, I was drawn to the clas- In fact, deeper analysis reveals beyond a doubt how any clas- sification work in the sciences, not because I thought it was sification, no matter how immutable it may seem, is after all worthier of being prioritized, but because it is linked to un- an artifact of the human imagination and functions in a par- derlying theories or conceptual frameworks. This link to the- ticular place and time. For many phenomena there are many ory seemed a very important one to me since it rendered the ways to interpret, visualize and explain them, even my classification as somehow more substantive, more enduring. revered periodic table. The value of any one approach de- I have since modified this view, but at the beginning of my pends on many factors, including the context in which the career I assumed that the more theoretical a classification classification is being invoked. Putting up any one classifica- was, the better it was at representing knowledge. I agreed tion as ideal implies that classification schemes that do not with Kaplan (1963) who said that theories and models are a measure up are somehow inferior rather than simply differ- “symbolic dimension of experience as opposed to the ap- ent, or that they do not have a useful purpose. While many prehension of brute fact” (294). In the same way, I thought traditional classifications have strength and merit, I have 658 Knowl. Org. 46(2019)No.8 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process come to realize the exclusive embrace of elite classifications Hierarchies have predetermined, predictable, and sys- and classificatory structures may build in subtle assumptions tematic rules for association and distinction. There are “nec- and biases, making it more difficult to admit diverse perspec- essary” and “sufficient” conditions for when something tives. may belong to one class and be distinguished from another class. The rules attempt to use the most “essential” type of 3.0. Classification structures information for distinguishing one class from another, and thus, entities differ from sibling entities in a predictable way. When I first investigated classification as a research topic the Often the rules are based on some theory that is the foun- question I posed was, “How do we systematically evaluate a dation for the classification. Finally, a hierarchy invokes the classification?” I wanted to step back and impartially analyze rule of “mutual exclusivity,” which means that each entity what made a classification tick—any classification. What are can belong to one and only one class. the features that provide strength in terms of knowledge To take one example, in western medicine, concepts lend representation, and what features misguide or interfere with themselves to hierarchical arrangement. Here’s an abbrevi- it? I was additionally motivated by my need to develop tech- ated snippet from the National Library of Medicine’s Medical niques for teaching about classification in a way that gave Subject Headings (MeSH) (https://meshb.nlm.nih.gov/rec students a toolkit they could use to describe and evaluate any ord/ui?ui=D013494). classification that came their way in a thoughtful and careful way, especially the legacy tools they would be using in their Nervous System Diseases [C10] work. An important part of this profile was a description of Central Nervous System Diseases [C10.228] the structural properties of the classification, how the parts Brain Diseases [C10.228.140] worked together in terms of relationships. There are many Basal Ganglia Diseases [C10.228.140.079] kinds of classificatory structures, but for this review I will Basal Ganglia Cerebrovascular Disease discuss the two that are probably most familiar: hierarchies [C10.228.140.079.127] and trees. Chorea Gravidarum [C10.228.140.079.294] .... 3.1 Hierarchies Parkinsonian Disorders [C10.228.140.079.862] Supranuclear Palsy, Progressive I start with hierarchies, because they are perhaps the most [C10.228.140.079.882] recognizable (and perhaps the most misunderstood). People Tourette Syndrome [C10.228.140.079.898] call all kinds of things hierarchies, but here I refer to the logical structure we have inherited from Aristotle (1963). In this example, searching for the rare disease progressive His view was that, after careful observation, one could learn supranuclear palsy, we learn that it is part of several linked how things could be aggregated and differentiated “natu- hierarchies. One is shown here, under nervous diseases. rally.” It assumed the division and aggregation of classes The information flows in many directions: from the top would be valid, because one had to arrive at the “essential terms down, from the “sister” terms, and also from the qualities” of what was being classified. Yes, we now question related terms (such as movement disorders) in other parts that such structures require us to choose one ideal represen- of the schedules. It helps the searcher identify the land- tation over possible others, but nevertheless, hierarchies are scape. Knowing that PSP is located near parkinsonian dis- often sought out as the structure of choice for their many orders and tourette syndrome helps define the nature of strengths described below. the disease, and indeed, helps explain why it is difficult to Essentially, a hierarchy is a structure with a top class that diagnose in its early stages. defines the scope of the classification. The top class in- It is obvious why such careful constructions are appeal- cludes all its subclasses and sub-subclasses. A true hierarchy ing in knowledge representation. If valid in their underly- has only one type of relationship, the “species/differentia” ing assumptions and definitions they offer complete and relationship, also known as the “generic” relationship, or in- comprehensive information. The affordance of inher- formally the “is-a” relationship. The strict requirement for itance provides for an economy of notation, and perhaps inclusion ensures that what is true for the top class is true most important, a hierarchy offers the ability to make in- for all the subclasses. This property is called “inheritance,” ferences from incomplete information. For instance, one that is, attributes are inherited by a subclass from its super- can infer that a female kitten is, like all cats, a mammal, and class. “Transitivity” is an important outcome of this care- by her essential mammalian features could be expected to fully controlled structure, because one can assume that all eventually bear live young and breast feed them, even if classes are members of not only their immediate superclass she is not at present doing so. Put another way, the hierar- but of every superclass above that one. chy succinctly carries a great deal of information that can Knowl. Org. 46(2019)No.8 659 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process be used to represent the domain, to explore, and to pro- sarily generic. There are many types of relationships pos- vide conceptual fodder for further discovery. Above all, a sible, such as part/whole, a kinship tree, or an organiza- hierarchy is built on logical principles, so to many people tional chart. In a part/whole scheme such as: it seems trustworthy. The parts must fit together; it should be comprehensive and not have loose ends. Building and North America → United States → NY State → maintaining hierarchies requires a strong à priori conceptual Onondaga County → City of Syracuse framework as well as consensus to guide the development of the rules. That is why hierarchies are often deductively you can see that Syracuse is a part of Onondaga County, created, rather than built from the bottom up. which is part of NY State, and so on, but Syracuse is not Not all knowledge domains lend themselves to a strictly a “kind of ” county, nor is the county a “kind of ” state. controlled hierarchical representation, though. Hierarchies This structure limits the transitivity of information, be- are problematic for a number of reasons. First and fore- cause what is true of North America does not carry down most, we acknowledge that it is often very difficult to iden- to the city level in terms of shared “essential” features. tify the ideal “essential” partition points in any domain of Similarly, if an engine of a car contains pistons, spark knowledge. Many phenomena can be seen from several per- plugs, and valves, you do not have a great deal of infor- spectives, depending on the context and the goal of the clas- mation about the relationship between pistons and spark sification. There may be multiple and diverse criteria, and plugs except to know they are both part of a car engine. there may be a question of which criterion to invoke first. They can, in fact, be totally different entities. What unifies For example, the traditional classification of animals divides them in the scheme is their position in the engine. them up into ever more specific taxonomic ranks, from In a traditional kinship tree, you may describe the flow kingdom to species. This order precludes the consideration of who begat whom, but a daughter is not a kind of of the differences between animals in the wild and animals mother, nor is a mother a kind of grandfather. Instead, the in captivity. Moreover, it is awkward to use this classification relationships are determined by blood and legal affiliation. to represent ecological systems gracefully, since there are In an organizational chart there is a clear purpose, and that many other factors to consider besides the animal’s morpho- is to show “who reports to whom” or perhaps “who is logical “essence.” managed by whom.” This is not to say that kinship trees Finally, a true hierarchy requires deep knowledge and or organizational charts do not yield a great deal of infor- consensus about the domain in order to determine the rules mation, but they are not as rich and inclusive as the classic for defining classes, partitioning, and aggregation. If there hierarchical structure in terms of showing the unity of the is no conceptual framework guiding these choices, the clas- whole. Instead, one or two critical relationships are high- sification can seem incoherent or contradictory. Thus, in lighted, which makes them easy to comprehend and ana- new and emerging fields where knowledge is incomplete, it lyze along those relationships. is sometimes unwise to commit to a hierarchical classifica- In summary, both hierarchies and trees are useful sys- tion. Even when the field is mature, though, but rapidly tematic knowledge-representation structures with differ- changing due to new incoming information, maintaining a ent properties and strengths and different constraints. In hierarchy can be dicey. For instance, we are familiar with the both you must know about the domain to pre-determine muddy classification of heavenly bodies, such as planets. the first cut and the rules for membership in classes. Both The principle of transitivity and inheritance requires that kinds of classification structures have many challenges in all the entities in a hierarchy be at the same level of concep- being applied, though, because the realities of application tual granularity, thus a hierarchy does not accommodate dif- do not always map well onto the requirements of such ferences of scale for the same phenomenon. For example, a structures. beach might be construed as a kind of land mass as you might see it from space, or it might be a kind of habitat, or 3.3 Facetted classification it might be viewed as an aggregate of materials, such as sand. A hierarchy encompassing all the views in one structure My growing awareness that classificatory thinking was would be impractical and confusing in terms of making in- both culturally and psychologically influenced led me to ferences and comparisons. further explorations. Among these was facetted classifica- tion, which is not a different representation structure, but 3.2 Trees rather an approach that allows the classifier to view the world as dynamic, and indeed, provisional, in how it is con- Trees are another familiar type of classificatory structure. strued. The approach is credited by many to S.R. Ranga- A tree divides and subdivides its classes just as in a hierar- nathan, an Indian scholar, who posited that any entity chy, but the relationship among the classes is not neces- could be viewed from a number of fundamental perspec- 660 Knowl. Org. 46(2019)No.8 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process tives or facets. He suggested that these are: personality, to have a required binding conceptual framework, this also matter, energy, space, and time (Ranganathan, 1967). means that the scheme remains essentially descriptive. While the discussion of the nature of these facets varies in There is no guidance for how to read or interpret the rela- his work, the principle has caught on and endured. The tionship “among” the facets. Finally, it may be difficult to notion that any entity can be analyzed into aspects, each visualize all in one grand picture. A scheme might include representing some feature or quality, freed up designers to a timeline, a hierarchy and a tree, but no built-in clue on create schemes that were multidimensional. Note this is how these should be presented. Even so, chances are if not the same as breaking down an entity into components, you look around at modern classification, it will likely as but rather viewing the same entity from different perspec- not be a facetted scheme (Kwaśnik 1999, 39-43). In much tives—same object different views. One of the clearest ex- of my research I’ve employed the notion of facetted clas- amples that explicitly built on Ranganathan’s principles is sification, especially in the challenges presented by multi- the Art and Architecture Thesaurus, a compilation of vocab- dimensional situations. ulary for the indexing and retrieval of objects and literature on material culture, which in its diversity defies easy de- 4.0 Practical classifications scription and classification into any one classificatory structure. The A&AT allows the creation of a string to The study of formal classification led me to an appreciation represent a topic or object using the core categories of pe- of their formidable power, but also piqued my interest in riod/style, place, process, material, and object. For exam- simple, practical classifications that perhaps deviated from ple: the “ideal.” Many of these exist to help with tasks such as shopping, diagnosis, or description without being neces- 19th Century Japanese raku ceramic vase sarily bound by a particular theoretical framework. That is, Arts & Crafts American oak desk they are not without an underlying rationale, but they do not purport to be “true” or enduring. They are simply there to In doing so it is then possible to search by any one of the organize some phenomenon in a useful way. After having components (e.g., all things “arts & crafts,” or all “vases”), scoffed at such schemes I became fond of them, because or in combination. It also allows for the graceful addition they demonstrate our human ability to be in touch with the of new objects and topics, so long as they can be analyzed power of a good classification that uses visualizations and using the five categories. simple metrics to help navigate through more complex in- This approach has extended well beyond formal collec- formation. The point of the following examples is that in tions and is very popular in shopping sites as well as visu- many ways they provide for enhancement of description alized analyses of all kinds. Not all use Ranganathan’s prin- and searching where sometimes the formal classifications ciples, but the result is essentially the same. Each facet can are lacking or overly complicated. These classifications be developed following its own logic and structure, and demonstrate that iterative, flexible design, combined with then synthesized into expressive strings. Obviously, the ad- other search features can be very effective and certainly eas- vantages are you do not need exhaustive knowledge so ier to maintain. long as you can identify important fundamental “aspects.” It is a hospitable and flexible approach without the need 4.1 Keys to have a strong, immutable theory for the scheme overall. At the same time, it can accommodate a variety of theo- My trusty old Peterson’s field guide to wildflowers (1974) retical structures and models in the facet components, and, is an example of such a pragmatic classification that makes most important, it can sustain a variety of perspectives. identification of wild plants more accessible. It is a key, Thus, a flower can be considered as food, as a feature of which is a type of classification that chooses one or two specific habitats, as a commodity, and so on. obvious features to lead into the more formal classifica- While facetted schemes have pragmatic appeal, there tion. In this case, the plants are organized by petal color, a are some things to consider. First, is the difficulty in iden- feature not “essential” in the Aristotelean sense but recog- tifying the core facets. They should be robust and so, while nizable. Then, by icons and clear descriptions, the identifi- complete knowledge of the domain is not necessary, cation can proceed further. In other words, the key is just enough must be understood to accommodate all new en- that, a key. The design of such a key might be frustrating tities. For example, you might view the traditional classifi- to the botanist because it might seem superficial, but to a cation of instruments as a facetted scheme, incorporating novice it is a testament to the communicative power of material, process of creating sound, origin, and so on, but such tools. the scheme hits a speed bump when you want to also in- clude electronic instruments. Second, while it is freeing not Knowl. Org. 46(2019)No.8 661 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process
4.2 eBay.com time, WebMD.com may benefit from some form of vo- cabulary control for richer expansion of terms and ar- Commercial websites use classifications in contexts where chival retrieval (Kwaśnik and Flaherty 2010). the content may be in constant flux, the user population is unknown, or if it is known, we can assume it is diverse; 5.0 Challenging the Aristotelian paradigm and where it is desirable that the classification be very sim- ple and straightforward so that all levels of users can learn As mentioned, two foundational but possibly conflicting as- it. For instance, eBay.com maintains a classification of mil- sumptions took shape in my early pursuit of studying clas- lions of objects. It fails miserably when analyzed using the sification: the first was my firm belief that formal classifica- formal criteria of coherence, but it is a classification of tory structures, such as hierarchies and trees, help advance current objects—everything on eBay exists and is for understanding because of their ability to represent not only sale—it is not meant to endure forever. Despite its rather the elements of a domain but also the relationships among sloppy design, it is surprisingly robust and hospitable. them, thus yielding knowledge structures that not only de- When you consider that it reflects the terminology used by scribe but also explain. At the same time, I realized that peo- several million people for an amazingly wide and con- ple in their enactment of classification brought their own stantly shifting array of items, it is quite impressive. In contextual understanding to them and the two did not al- terms of accessibility, there are very few terms in the main ways map well. categories that are difficult to understand. One of the strong points is the meshing of the classification with 5.1 Accounting for context in the classification of many other access strategies. If the classification fails, personal documents there are other avenues to pursue (Kwaśnik and Liu 2000). In my dissertation, The Influence of Context on Classificatory Be- 4.3 Amazon.com havior (Kwaśnik 1989b, 1991), I wanted to see how people create and then utilize classification, that is, I wanted to learn Along the same lines, when I studied the amazon.com book about personal information management in everyday life. I division years ago (Kwaśnik, 2002), even back then the af- interviewed university professors in their offices and rec- fordances of multiple access points ensured that “some- orded their documents and the organization of these docu- thing” would be found no matter what the user entered. The ments on shelves, in drawers, on the computer, in their brief- classification achieved a multi-perspective view allowing for cases, taped to the door, and in various piles and files, always a facetted approach, and if one approach did not work an- in their own words. The findings, in a nutshell, were they did other was readily available. I concluded (284) that “in gen- not organize things as they are organized in formal library eral, amazon.com’s scheme can be viewed as more prag- collections by subject and form. In fact, the contextual fac- matic and enumerative than as based on a model of tors played a critical role—factors such as the purpose of knowledge.” It is a classification meant to encourage buying the document or its currency. Documents with the same and uses as many routes to the goal as possible, including a content or subject could be classified differently depending simple but redundant vocabulary without much attention to on how they would be used (Kwaśnik 1989a). The findings structural integrity but able to provide a rich network of suggested that while formal classifications exist and prove subjects. very useful, the establishment of universal schemes is more problematic in situations in which context plays a part. So, 4.4 Scientific and naïve classifications formal logical representation works when the domain is well understood and there is consensus on the underlying con- For a final example, we explored the idea of “teaming up” ceptual structure, but what about all the rest? scientific and naïve classifications. We compared two sep- arate but related classification schemes in the area of med- 5.2. Influences of cognitive anthropology ical information to better understand how they might be used together and inform one another. We contrasted Along the way to finding a conceptual framework for my MeSH with the consumer health website, WebMD.com. dissertation I was introduced to the field of cognitive an- Using the term “autism” we compared the strengths and thropology, and most intriguing to me, category choosing. limitations from the perspective of vocabulary, syntax and In a hierarchy, each member of a class is an equally good classificatory structure, context, and warrant. We conclude representative of that class, since each member must pos- that in terms of vocabulary and concepts, MeSH may ben- sess all the requisite attributes. In a pure hierarchy an entity efit from WebMD’s approach to ongoing updates and cur- cannot be “sort of ” in a class. The boundaries and rules rency as well as the contextualization of terms. At the same for inclusion and exclusion are defined and predetermined. 662 Knowl. Org. 46(2019)No.8 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process
An entity cannot belong to more than one class following ing of office documents. The study yielded a fine-tuned the principle of mutual exclusivity. These traditional as- descriptive analysis of consensus, conflict, and corre- sumptions came increasingly into question, because an- spondence among people for common documents, thropological and cognitive evidence from studies of hu- demonstrating that in fact perfect correspondence in nam- mans did not support them. The formal properties do not ing between individuals is not the norm (Kwaśnik and Jör- necessarily map accurately to our human cognitive pro- gensen 1992). cesses, that is, we do not all store our concepts in hierar- chies. We know that humans have fuzzier notions of what 6.0 Extending the borders constitutes a boundary on a class of things. These bound- aries may change with circumstances and the experiences In the time I have been studying classification, we have of the classifier. seen a shift to unification and standardization of biblio- graphic systems, not just in the United States but also glob- 5.2.1 Prototype theory and the principle of family ally. This means that traditional classifications, originally resemblances designed in a particular country or for a particular collec- tion are now being stretched, in Michèle Hudon’s words I was influenced in particular by researchers such as Eleanor (1997), to cover cultural and linguistic artifacts and con- Rosch whose prototype theory posits that not all members cepts quite different from those originally intended. This of a class may be perceived as equally good representatives had special significance for me, because given my under- of that class. Not all birds fall naturally into the class of standing of a classification’s structure and impact, I knew “birds;” some seem more birdlike and others less so. She that extending them was not simply a matter of one-to- posited that for every class, some objects become prototyp- one translation. ical and form a best example of that class (Rosch 1973). An- other notion that influenced me was the idea of family re- 6.1 Translating classifications semblances (Rosch and Mervis 1975), where they argue that the principle of family resemblance can be construed as “a In a study comparing the Dewey Decimal Classification and logical alternative to criterial attributes.” They were arguing the Korean Decimal Classification, two bibliographic against (603) “a tenacious tradition of thought in philoso- schemes from different cultures, we found that obvious phy and psychology which assumes that items can bear a differences could be accommodated (Kwaśnik and Chun categorical relationship to each other only by means of the 2004). For instance, the DDC emphasized Christianity, possession of common criterial attributes.” Their study pre- while the KDC allowed more room for Buddhism. The sented empirical confirmation that formal criteria are nei- KDC offered greater expressiveness for terms such as ther a logical nor psychological necessity. This means that “calligraphy.” The differences that were more profound, for any given class, certain attributes define members of that however, were those that construed subjects very differ- class, but not all members must possess all the attributes nor ently. For example, “war” is treated as a social process in exhibit them as strongly (thus defying the principle of nec- the DDC, and is placed near diplomacy, whereas in the essary conditions). Imagine in your family there are recog- KDC it is classified as a social problem and is near suicide nizable family traits, but they are not distributed equally (197). Such a difference in conceptual mapping makes cul- among everyone. The idea of family resemblances raised turally sensitive translation challenging. some interesting questions, such as which of these attributes Translating the vocabulary of a classification has the is defining? Are all combinations defining? typical issues of translation in general. In a study of peo- ple’s use of even the very most basic kinship terms such as 5.2.2 George Kelly and personal construct theory “mother,” there are many problems. Among these are find- ing corresponding terminology and being able to reflect Another important contribution to understanding the cog- the relationship between terms in the target language cor- nitive aspects of classification was George Kelly’s Personal rectly. It is surprising how many denotations and connota- Construct Theory (1955, 1970). Kelly posited that every- tions the term “mother” has, even in English. We found in one construes the world in a different and individual way. the process of translation there may be structural shifts; His original work included an intriguing appendix: The some terms may have broader definitions and some nar- Repertory Grid. This tool eventually was used outside its rower. There may be differences in how similar terms are original intention and became popular for making people’s construed, and there may be additional criteria of distinc- individual implicit constructs explicit. Building on the find- tion (such as birth order in the case of kinship systems). ing from my dissertation that variations in naming were We suggested that not only terms themselves but also in- large, we used the Repgrid technique to explore the nam- ter-term relationships need to be preserved in cross-cul- Knowl. Org. 46(2019)No.8 663 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process tural, cross-lingual classification translations so that both would provide a set of principles for designing browsable the source and the target schemes are truly reflective interfaces. (Kwaśnik and Rubin 2003). The studies showed that with respect to the structure of the environment, the notion of an unstructured environ- 7.0 The importance of context ment is probably not as useful as observing what structures are perceived and how they affect behavior. People will cre- It was evident to me that a key ingredient in making clas- ate structures even amidst seeming chaos. Similarly, they will sifications more responsive and resonant was to find some develop a purpose to the process, even if they seemingly way of incorporating context into the process. A professor started off without one. Comparisons and strategy are de- organizes office documents with an eye to the potential veloped iteratively. This amazing human capability can be uses. A person browsing a collection brings to it personal described by the following functions: orientation, place insights or needs and uses those to help navigate the space. marking, identification, resolution of anomalies, compari- The situation in which classificatory decisions are made sons, and transitions. Each of these functions is performed plays an important part. Yet, it is quite difficult to reconcile by constant interaction with the browsing environment, but rigid classification schemes with infinitely individual ones. also with past experiences, future plans, and many other fac- Starting with the findings of my dissertation, the dilemma tors the browser brings to the experience. Thus, we can say of creating and using classifications that are accommodat- that browsing is not a passive activity, because there is a for- ing of many perspectives always seems to boil down to one midable amount of sense making involved. As a way of cop- important factor: context. Context defines the scope and ing with the browsing environment, the browser is con- the vocabulary. It decides on the elements themselves and stantly devising classifications or views based on a shifting which classificatory relationships are pertinent. The fol- context. Being able to harness these abilities would make in- lowing examples show two streams of research in which I terfaces easier and more productive (Kwaśnik 1993). tried to find ways of identifying and then representing contextual factors. 7.2 Genres
7.1 Context and discovery: browsing My growing awareness that classification of any kind is a social act led me to the study of genres as they play out in One of the features of a classification, any classification, is knowledge representation for information seeking and that it creates affordances for exploration. A classified envi- use. A genre identifies something as an integrated cluster ronment can be searched or browsed for something even if of features enacted in a social environment. My colleague, we only suspect or expect it will be there but do not know Kevin Crowston, and I conducted a series of studies to see for sure. Browsing is a method of information seeking that if identifying the genre of a document would improve in- allows the user to explore and navigate without having to formation access in large digital collections through the specify a query. As such, it is a good way of dealing with an identification of document genre as a facet of document unfamiliar environment or with multiple options or choices. and query representation. For instance, knowing some- In this way, browsable systems can be invaluable to users thing was a computer program might help distinguish it crossing over into a new and unfamiliar domain. Browsing from a musical program, each of these being a different reduces cognitive load, because it is generally easier to rec- genre. Because most genres are characterized by both form ognize something once it is viewable rather than to recall a and purpose, identifying the genre provides information as term for it. As well, a key feature of browsing is the ability to the document’s purpose and its fit to a user’s situation, to hold several parallel paths at once without having to com- which can be otherwise difficult to assess (Crowston and mit to just one. Kwaśnik 2003, 2005). I wanted to investigate what people do when they First, we needed to define genres for ourselves since this browse. The term had been variously defined as searching, is a very old area of study and crosses many disciplinary lines, scanning, navigating, skimming, sampling, and exploring. It from the arts to business. Genres are a way people refer to was often described as searching “without a particular pur- communicative acts that is understood by them, more or pose” and without a set structure as compared to a database less, but is often difficult to describe in its particulars. Thus, search, for instance. We conducted some informal observa- genres are recognized and used, but not so readily described tions of people browsing in catalogs, online, at a farmer’s and defined. In our work, we drew on the definition of market, and so on (Kwaśnik and Yoon 1990). The purpose genre proposed by Yates and Orlikowski (1992, 543), who was not simply to record what they did or what “nodes” they describe genre as “a distinctive type of communicative ac- visited and how often, but more fundamentally to identify tion, characterized by a socially recognized communicative what function they accomplished. Ultimately, we hoped this purpose and common aspects of form.” Note this does not 664 Knowl. Org. 46(2019)No.8 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process mean that a genre can be seen purely as a set of document of South Africa’s history, or the classification of tubercu- attributes, making the representation of genres a complex losis affected people’s life trajectories. and difficult proposition. Classification schemes reflect the knowledge of the do- Among other things, we wanted to know how people main being classified but also the perspective of the clas- talk about the genre of documents. How do people make sifier, thus no classification can ever by understood out of use of new, unnamed, and emerging genres? What clues context. While we take for granted that classifications do do people use to identify genre when engaged in infor- have a social impact, it is not always easy to say precisely mation-access activities? What facets (basic attributes) of how, although we can certainly feel the effects. Potential genres do people perceive (Crowston and Kwaśnik 2003)? marginalization, rules for inclusion or exclusion, labeling Our plan was to create a taxonomy of genres by studies of and naming are all outcomes of classification decisions. people searching for information in the field. This taxon- Those in power design the classification and then have omy would be used to create a simulated search situation power over those who are not able to change it. The news in which we could observe the difference between search- is full of examples on a daily basis, from pressure on the ing with the aid of genre information and without. Library of Congress to change the term “illegal aliens” to Our assumption going in was that a facetted scheme for “undocumented immigrants” to who can use the term genres would be best given their multidimensionality and “champagne.” Political resistance often means changing complexity (Kwaśnik and Crowston 2004). We attempted the ruling classification. Many standards are based on clas- to harvest clues people used to identify genres, such as sification. Many conflicts have at their core a dispute over “scholarly language” or “reverse chronological dated con- basic classifications: when does life begin and when does tent” and then reduce them by analysis into possible genre death occur? What makes a crime a crime? What defines a facets. The clues and resulting user-generated scheme were country? In learning how to evaluate a classification we not possible, because the concept of genre was even more should always take the critical view. Who devised it? Whose slippery than we anticipated. We had difficulty defining purpose is being served? genres and developing the scope and expressiveness of the scheme from what our participants told us. They were not 8.1 The case of ontological commitment and warrant able to reliably identify the genre unit or provide unambig- uous genre labels. When prompted, they found it difficult Sorting Things Out engendered a critical eye with respect to to identify genre attributes. There were challenges in dis- my analysis and perception of classification, but it was one tinguishing form and content, as well as challenges in iden- thing to find the strengths and flaws and another to develop tifying purpose. Finally, the granularity of their tasks dif- a vocabulary for discussing this systematically and coher- fered immensely creating imbalances in the granularity of ently. Fortunately, I was asked to contribute to a festschrift the terms we could use (Kwaśnik, et al. 2006). As a result, for Claire Beghtol (Kwaśnik 2010) and chose to focus on we worked around the lack of a user-generated facetted her pivotal and far-reaching 1986 article “Semantic Validity: view of genres and created a researcher-compiled working Concepts and Warrant in Bibliographic Classification Sys- taxonomy for the purposes of the experiments (Crowston, tems” (Beghtol 1986). In this article she explores the seman- et al. 2011). In the end, we were not able to demonstrate a tic, rather than the syntactic axis of bibliographic classifica- substantive change between plain searches and those en- tion systems. According to her, the attention of scholars on riched with genre information, but I believe the full poten- facetted schemes and classificatory structures had hereto- tial of genre representation remains to be explored. fore pulled our attention to the syntactic aspects (e.g., con- cept division and citation order), with semantics being con- 8.0 Classification at the intersection with human sidered more or less a question of the terms and their rela- endeavor tionships and somewhat taken for granted. In this paper she states (110-11) that “the warrant of a classification system Sorting Things Out by Geoffrey C. Bowker and Susan Leigh can be thought of as the authority a classificationist invokes Star (1999), made an enormous impression on me and on first to justify and subsequently to verify decisions about the knowledge organization community and beyond. In what class/concepts should appear in the schedules …. The this work, the authors examined revelatory classifications semantic warrant of a system thus provides the principal au- and standards to show how such classifications silently in- thorization for supposing that some class or concept or no- fluence the infrastructure of information, affecting not tational device will be helpful and meaningful to classifiers only policy, but also our daily lives. They were not the first and ultimately to the users of documents.” Warrant emerges to urge that we question classifications, but their insights from various points of authority: literary warrant, scien- were profound, vivid, and compelling. They showed how tific/philosophical warrant, educational warrant, and cul- the system of apartheid, for instance, embodied the pain tural warrant, each with its own effect in terms of establish- Knowl. Org. 46(2019)No.8 665 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process ing the semantics and then also the syntax of any given clas- My takeaway from studying these cases is that, broadly sification (119-221). speaking, when classification is structured to support hu- This was a revolutionary idea in the sense that notions man endeavors, the purpose is different than when it is of meaning being fixed have guided the design of many structured to support science. Thus, understanding the un- of our systems, because it was assumed that meaning be- derlying warrant is all the more important. came more stable and consensus firmer as the evidence mounted and the ideas withstood the test of time. Yet, 9.0 Full round back to hierarchies modern approaches assume that meaning is not fixed and is created in use. It is also interesting to consider contem- In a way, then, my early respect for hierarchies would seem porary phenomena such as wikipedia.org, where the clas- to have been validated up to a point, if, and only if, the sification and the content are built cooperatively. That is, circumstances supporting a hierarchical structure were ev- in principle, both the text and the classification that organ- ident. Recently, though, even this qualified view has been izes the texts in such emergent systems are not managed somewhat shaken. Hope Olson’s mission is to analyze our from the top. Nobody questions the fact that such systems traditional knowledge representation systems from the must be flexible and dynamic, and yet nobody wants an point of view of those whose voices are not well reflected. amorphous mess either. Our challenge is to assess the war- In her article “How We Construct Subjects” (2007), she rant for any given classification project and judge the clas- takes apart the notions behind hierarchies and brings to sification against it. bear feminist thinking to offer a penetrating critique. She One example of such a challenge is the classification of posits that hierarchies are by nature flawed because they academic departments and programs. In the modern require one element to be in the superior position and all American university, there is often a federated system of other elements subordinate to that. This structure creates schools and colleges, each with its own warrant for how it skewed assumptions that privilege one set of elements describes and labels the knowledge in its purview, what over others. I will use my own example here: imagine a hi- Elaine Svenonius called ontological commitments (Sveno- erarchical classification of astronauts. At the top is the nius 1997). Chemistry views its own world differently than term “astronauts.” On the next level down are subclasses do the performing arts, and the differences are evident in of astronauts: “minority astronauts,” “women astronauts,” academic processes. In a study of my own university, I and so on. This may seem like a laudable effort at inclu- used the collection of hundreds of courses to see how this siveness. There is no subclass for “men astronauts,” how- played out (Kwaśnik 2016). It is clear, for example, that the ever, because the notion of astronauts being men is the term “girl education” is construed differently in econom- default and is baked into the assumptions. In the chain of ics than it is in education. In one the presence of girl edu- transitivity, men hold the defining set of attributes. This is cation is a factor in economic development, and in the a dilemma, because while one would like a way to represent other a subject of interest in its own right. Resolving such the special attributes of women astronauts, placing them territorial disputes in claiming courses is left to the curric- in the subordinate position means that they are defined by ulum committees. the male criteria first and foremost. From my perspective, A more interesting example, though, is forensic science, there does not seem to be a good way to undo this imbal- an instance of mixed ontological commitments. There are ance in a hierarchy. many such examples at universities: archival studies, phys- Having laid out the limitations both in content and ical education, and environmental studies among them. structure, Olson suggests rewriting and restructuring our The study of forensic science is the use of science to help schemes so that the all-important connections are visi- solve crimes. It calls upon an array of disciplines to sup- ble—a web instead of a hierarchy (522). According to her port a specified set of professional practices. Courses in (522), we need “richer and more situated logical models” the FS curriculum include forensic anthropology, human that allow for the representation of interdependence and osteology, forensic entomology, forensic chemical analysis, connectedness. I am just now beginning to rethink how my forensic linguistics, forensic evidence in law, and forensic favorite classifications could be reframed in this way, or if psychology. Each of these comes with its own ontological they even should be. The power of hierarchies and other commitments and its own body of knowledge. The foren- formal classifications is not easily dismissed, but at the sics student’s program of study is not based on the sup- same time the fact that they are so embedded in our culture porting and contributing disciplines, however, but rather should be explored. What we see as taken for granted on a prescribed sequence of professional practice: identi- could be hiding subtle and not so subtle biases. fication of crime; collection of evidence (autopsy, traces); analysis of evidence; and support of the preparation of a legal case. 666 Knowl. Org. 46(2019)No.8 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process
9.0 Summary national ACM SIGIR Conference on Research and Development in Information Retrieval Cambridge, Massachusetts, USA June Classification is beautifully recursive. What we know guides 25 – 28, 1989. New York: ACM, 207-10. doi:10.1145/ our classifications, and in turn, our classifications guide what 75334.75356 we are able to know. Many questions remain: 1) Who creates Kwaśnik, Barbara H. 1989b. "The Influence of Context the classifications by which we must all live?; 2) Who has the on Classificatory Behavior." PhD diss., Rutgers. authority to change them?; and, 3) What is an effective way Kwaśnik, Barbara H. 1991. The Importance of Factors of creating classifications that are inclusive but also effec- that Are Not Document Attributes in the Organization tive? We use classifications to better capture what we know; of Personal Documents.” Journal of Documentation 47: we also use them to embody our values and perspectives. We 389-98. don’t have a choice of whether to classify or not, but we are Kwaśnik, Barbara H. 1993. “A Descriptive Study of the obliged to pay attention to the consequences of what we do. Functional Components of Browsing.” In Proceedings of the IFIP TC2/WG2.7 Working Conference on Engineering for References Human-Computer Interaction Ellivuori, Finland, August 10- 14, 1992, ed. James A. Larson and Claus Unger. Am- Aristotle. 1963. Categories, and De interpretatione, trans. J.L. sterdam: North-Holland Publishing, 191-203. Ackrill. Clarendon Aristotle series. Oxford: Clarendon Kwaśnik, Barbara H. 1993“The Role of Classification Press. Structures in Reflecting and Building Theory.” In Proceed- Beghtol, Clare. 1986. “Semantic Validity: Concepts of War- ings of the 3rd ASIS SIG/CR Classification Research Work- rant in Bibliographic Classification Systems.” Library Re- shop, ed. Raya Fidel, Barbara H. Kwaśnik and Philip sources & Technical Services 30: 109-25. Smith. Advances in Classification Research 3. Medford, Bowker, Geoffrey C. and Susan Leigh Star. 1999. Sorting NJ: Learned Information, 63-82. Things Out: Classification and Its Consequences. Cambridge, Kwaśnik, Barbara H. 1999. “The Role of Classification in MA: MIT Press. Knowledge Representation and Discovery.” Library Bronowski, Jacob. 1978. The Origins of Knowledge and Imagi- Trends 48, no. 1: 22-47. nation. New Haven, CT: Yale University Press. Kwaśnik, Barbara H. 2002. “Commercial Websites and the Crowston, Kevin and Barbara H. Kwaśnik. 2003. “Can Use of Classification Schemes: The Case of ama- Document-Genre Metadata Improve Information Ac- zon.com.” In Challenges in Knowledge Representation and Or- cess to Large Digital Collections?" Library Trends 52, no. ganization for the 21st Century: Integration of Knowledge Across 2: 345-61. Boundaries: Proceedings of the Seventh International ISKO Con- Crowston, Kevin, Barbara H. Kwaśnik, and Joseph Ru- ference, 10-13 July 2002, Granada, Spain, ed. Mariá J. López- bleske. 2011. “Problems in the Use-centered Develop- Huertas; and Francisco J. Muñoz-Fernández. Advances ment of a Taxonomy of Web Genres.” In Genres on the in Knowledge Organization 8. Würzburg: Ergon, 279- Web: Computational Models and Empirical Studies, ed. Alex- 85. ander Mehler, Serge Sharoff, and Marina Santini. Dor- Kwaśnik, Barbara H. 2010. “Semantic Warrant: A Pivotal drecht: Springer, 69-84. Concept for Our Field.” Knowledge Organization 37: 106- Davies, Roy. 1989. “The Creation of New Knowledge by 10. Information Retrieval and Classification.” Journal of Kwaśnik, Barbara H. 2011. “Approaches to Providing Con- Documentation 45: 273-301. text in Knowledge Representation Structures.” In Classi- Hudon, Michèle. 1997. “Multilingual Thesaurus Construc- fication & Ontology: Formal Approaches and Access to tion: Integrating the Views of Different Cultures in Knowledge: Proceedings of the International UDC Seminar 19- One Gateway to Knowledge and Concepts.” Knowledge 20 September 2011, the Hague, the Netherlands, ed. Aida Organization 24: 84-91. Slavic and Edgardo Civallero. Wurzburg: ̈ Ergon, 9-24. Kaplan, Abraham. 1967. The Conduct of Inquiry: Methodology Kwaśnik, Barbara H. 2016. “The Web and the Pyramid: for the Social Sciences. Addison-Wesley. Hope Olson’s Vision of Connectedness in a World of Kelly, George A. 1955. The Psychology of Personal Constructs. Hierarchies.” Knowledge Organization 43: 367-72. New York: Norton. Kwaśnik, Barbara H. and You-Lee Chun. 2004. “Translation Kelly, George A. 1970. “A Brief Introduction to Personal of Classifications: Issues and Solutions as Exemplified in Construct Theory." In Perspectives in Personal Construct the Korean Decimal Classification.” In Knowledge Organization Theory, ed. D. Bannister. London: Academic Press, 1-29. and the Global Information Society: Proceedings of the Eighth In- Kwaśnik, Barbara H. 1989a. “How a Personal Document’s ternational ISKO Conference, 13-16 July 2004, London, UK, Intended Use or Purpose Affects Its Classification in an ed. Ia C. McIlwaine. Advances in Knowledge Organiza- Office.” In SIGIR ‘89 Proceedings of the 12th Annual Inter- tion 9. Würzburg: Ergon, 193-8. Knowl. Org. 46(2019)No.8 667 B. H. Kwaśnik. Changing Perspectives on Classification as a Knowledge-Representation Process
Kwaśnik, Barbara H. and Kevin Crowston. 2004. A Frame- Kwaśnik, Barbara H. and Kyung-Hye Yoon. 1990. “A De- work for Creating a Facetted Classification for Genres: sign Study for the Investigation of Browsing Behavior.” Addressing Issues of Multidimensionality.” In Proceedings In CAIS/ACSI ‘90: Proceedings of the Canadian Association of the 37th Annual Hawaii International Conference on System for Information Science Annual Conference, Kingston, Ontario, Sciences: Abstracts and CD-ROM of Full Papers: 5-8 January, May 24-26, 1990. 2004, Big Island, Hawaii, ed. Ralph H. Sprague, Jr. Los Olson, Hope A. 1998. “Mapping beyond Dewey’s Bounda- Alamitos, CA: IEEE Computer Society Press. doi:10. ries: Constructing Classificatory Space for Marginalized 1109/HICSS.2004.1265268. Knowledge Domains.” Library Trends 47: 233-254. Kwaśnik, Barbara H. and Kevin Crowston. 2005. “Genres Olson, Hope A. 2007. “How We Construct Subjects: A of Digital Documents: Introduction to the Special Is- Feminist Point of View.” Library Trends 56, no. 2: 509-41. sue.” Information, Technology & People 18: 76–88. Peterson, Roger Tory and Margaret McKenny. 1974. A Field Kwaśnik, Barbara H. and Mary Grace Flaherty. 2010. Guide to Wildflowers of Northeastern and North-central North “Harmonizing Professional and Non-professional America: A Visual Approach Arranged by Color, Form, and Classifications for Enhanced Knowledge Representa- Detail. The Peterson Field Guide Series 17. Boston: tion.” In Paradigms and Conceptual Systems in Knowledge Or- Houghton Mifflin. ganization: Proceedings of the Eleventh International ISKO Rosch Eleanor H. 1973. “Natural Categories.” Cognitive Psy- Conference, 23-26 February 2010, Rome, Italy, ed. Claudio chology 4: 328-50. Gnoli and Fulvio Mazzocchi. Advances in Knowledge Rosch Eleanor H. and Carolyn Mervis. 1975. “Family Re- Organization 12. Wurzburg: ̈ Ergon, 229-35. semblances: Studies in the Internal Structure of Cate- Kwaśnik, Barbara H. and Corinne Jörgensen. 1992. “The gories.” Cognitive Psychology 7: 573-605. Exploration by Means of Repertory Grids of Semantic Ranganathan, S. R. 1967. Prologomena to Library Classification. Difference among Names for Office Documents.” In 3rd ed. Bombay: Asia Publishing House. Classification Research for Knowledge Representation and Or- Svenonius, Elaine. 1997. “Definitional Approaches in the ganization: Proceedings of the 5th International Study Confer- Design of Classification and Thesauri and Their Impli- ence on Classification Research, ed. Nancy J. Williamson and cations for Retrieval and Automatic Classification.” In Michèle Hudon. Amsterdam: Elsevier. Knowledge Organization for Information Retrieval, ed. L.C. Kwaśnik, Barbara H. and Xiaoyong Liu. 2000. “Classifica- McIlwaine. FID 716. The Hague: International Federa- tion Structures in the Changing Environment of Active tion for Information and Documentation, 12-16. Commercial Websites: The Case of eBay.com”. In Dyna- Yates, Joanne and Wanda J. Orlikowski. 1992. “Genres of mism and Stability in Knowledge Organization: Proceedings of the Organizational Communication: A Structurational Ap- Sixth International ISKO Conference, 10-13 July 2000, Toronto, proach to Studying Communications and Media.” Acad- Canada, ed. Clare Beghtol, Lynne C. Howarth and Nancy emy of Management Review 17: 299-326. J. Williamson. Advances in Knowledge Organization 7. Würzburg: Ergon, 372-7. Kwaśnik, Barbara H. and Victoria L. Rubin. 2003. “Stretch- ing Conceptual Structures in Classifications across Lan- guages and Cultures.” Cataloging & Classification Quarterly 37: 33-47.
668 Knowl. Org. 46(2019)No.8 Knowledge Organization. Sixth Annual “Best Paper in KO Award” for Volume 45 (2018)
Knowledge Organization
Sixth Annual “Best Paper in KO Award” for Volume 45 (2018)*
DOI:10.5771/0943-7444-2019-8-668
Awarded to:
Bullard, Julia. 2018. “Curated Folksonomies: Three Implementations of Structure Through Human Judgment.” Knowledge Organization 45(8): 643-652. 57 references. DOI:10.5771/0943-7444-2018-8-643.
Abstract: Traditional knowledge organization approaches Julia Bullard is an Assistant Professor at the School of In- struggle to make large user-generated collections naviga- formation at the University of British Columbia. She holds ble, especially when these collections are quickly growing, a PhD in information studies from the University of Texas in which currency is of particular concern, for which pro- at Austin, an MLIS from the University of British Colum- fessional classification design is too costly. Many of these bia, an MA in cultural studies & critical theory from collections use folksonomies for labelling and organization McMaster University, and a BA in english literature from as a low-cost but flawed knowledge organization ap- St. Francis Xavier University. She teaches courses in infor- proach. While several computational approaches offer mation organization, metadata, and social justice. Her cur- ways to ameliorate the worst flaws of folksonomies, some rent research projects focus on how terminology systems user-generated collections have implemented a human depict LGBTQ2IA+ subjects and express theories of na- judgment-centered alternative to produce structured folk- tional values. sonomies. An analysis of three such implementations re- veals design differences within the space. This approach, termed “curated folksonomy,” presents a new object of study for knowledge organization and represents one an- * Awards Committee for Volume 45 (2018): Michael swer to the tension between scalability and the value of Kleineberg, chair; Devika Madalli, Heather Moulaison human judgment. Sandy and Ann Graf. Knowl. Org. 46(2019)No.8 669 Books Recently Published
Books Recently Published Compiled by J. Bradford Young
DOI:10.5771/0943-7444-2019-8-669
Auber, Olivier. 2019. Anoptikon: Une exploration de l’Internet Greselin, Francesca, Laura Deldossi, Luca Bagnato and invisible: Échapper à la main de Darwin. Actions & re- Maurizio Vichi, eds. 2019. Statistical Learning of Complex cherches. [Limoges]: Fyp. Data. Studies in Classification, Data Analysis, and Bauer, Nadja, Katja Ickstadt, Karsten Lübke, Gero Sze- Knowledge Organization. Cham: Springer. pannek, Heike Trautmann and Maurizio Vichi, eds. Gurrin, Cathal, ed. 2019. 2019 International Conference on 2019. Applications in Statistical Computing: From Music Data Content-Based Multimedia Indexing (CBMI). [Piscataway, Analysis to Industrial Quality Improvement. Studies in Clas- NJ]: IEEE. sification, Data Analysis, and Knowledge Organization. Hall, Christopher J. and Rachel Wicaksono. 2019. Ontologies Cham: Springer. of English: Conceptualising the Language for Learning, Teach- Berlia ︠ vskii ︡ , ̆ L. G. and S.N. Danikhno. 2019. Klassifikats ︠ ii ︡ a︠ ︡ ing, and Assessment. Cambridge Applied Linguistics. New istochnikov konstitut s︠ ionnogo ︡ prava: Voprosy teorii i praktiki: York: Cambridge University Press. monografiia ︠ . ︡ Konstituts ︠ ionnoe ︡ pravo. Moskva: Izda- Handel, Lisa. Ontomedialität: Eine medienphilosophische Perspek- tel’stvo IU ︠ rlitinform. ︡ tive auf die aktuelle Neuverhandlung der Ontologie. Medien- Bronner, Simon J. 2019. The Practice of Folklore: Essays To- Kultur-Analyse 11. Bielefeld: Transcript. ward a Theory of Tradition. Jackson: University Press of Haynes, David and Judi Vernau, eds. 2019. The Human Po- Mississippi. sition in an Artificial World: Creativity, Ethics and AI in Cavallini, Graziano. 2019. Che cos’è la realtà: Significato di Knowledge Organization: ISKO UK Sixth Biennial Conference un’idea e di una parola. Canterano: Aracne. London, 15-16th July 2019. Baden-Baden: Ergon. Crowther, Paul. 2019. The Aesthetics of Self-Becoming: How Kendall, Elisa F. and Deborah L. McGuinness. 2019. On- Art Forms Empower. Routledge Research in Aesthetics. tology Engineering. Synthesis Lectures on the Semantic New York: Routledge. Web: Theory and Technology 18. [San Rafael, CA]: Darder, Antonia, ed. 2019. Decolonizing Interpretive Research: Morgan & Claypool. A Subaltern Methodology for Social Change. London: Kogachi, Ryūichi. 2019. Mokurokugaku no tanjō: Ryū Kyō ga Routledge. unda shomotsu bunka. Kyōdai jinbunken tōhōgaku sōsho Davis, Matthew Evan, Tamsyn Mahoney-Steel and Ece 6. Kyōto-shi: Rinsen Shoten. Turnator, eds. 2019. Meeting the Medieval in a Digital Meissner, Dennis. 2019. Arranging and Describing Archives World. Medieval Media Cultures. Leeds: Arc Humanities and Manuscripts. Archival Fundamentals Series III 2. Chi- Press. cago: Society of American Archivists. Day, Ronald E. 2019. Documentarity: Evidence, Ontology, and Ndubisi, Ejikemeuwa J. O., Jude I. Onebunne and Paul T. Inscription. History and Foundations of Information Sci- Haaga. 2019. Igwebuike Ontology: An African Philosophy of ence. Cambridge, MA: MIT Press. Humanity Towards the Other: Papers in Honour of Professor De Kosnik, Abigail and Keith P. Feldman, eds. 2019. #Iden- Kanu, Ikechukwu Anthony. Milton Keynes: AuthorHouse tity: Hashtagging Race, Gender, Sexuality, and Nation. Ann UK. Arbor, MI: University of Michigan Press. Peris-Ortiz, Marta, João J. Ferreira and Jose M. Merigó Lin- Deana, Danilo. 2019. A ciascuno il suo catalogo: La rivoluzione dahl. 2019. Knowledge, Innovation and Sustainable Develop- tecnologica e le biblioteche. Biblioteconomia e scienza dell’in- ment in Organizations: A Dynamic Capabilities Perspective. In- formazione 23. Milano: Editrice Bibliografica. novation, Technology, and Knowledge Management. Dīkshita, Pratibhā. 2019. Sat kā svarūpa: Advaitavādī evaṃ Cham: Springer. ādhunika pariprekshya meṃ. Dillī: Vidyānidhi Prakāśana. Phillips, Laura. 2019. Profiles of Collection Cataloging and Doc- Eldred, Michael. 2019. Movement and Time in the Cyberworld: umentation Efforts by Museums, Archives & Library Special Questioning the Digital Cast of Being. Berlin: De Gruyter. Collections. [New York]: Primary Research Group. Godby, Jean. 2019. Creating Library Linked Data with Wiki- Vukadin, Ana. 2019. Metadata for Transmedia Resources. base: Lessons Learned from Project Passage. Dublin, OH: Chandos Information Professional Series. Cambridge, OCLC Research. MA: Chandos.