KO KNOWLEDGE ORGANIZATION Contents

Knowl. Org. 44(2017)No.2

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

Errata...... 83 Gems from our Digitization Project

Articles Douglas J. Foskett. Systems Theory and its Relevance Xiao-Bo Tang, Wei Wei, Guang-Chao Liu to Documentary Classification...... 129 and Juan Zhu. An Inference Model of Medical Insurance Books Recently Published ...... 135 Fraud Detection: Based on Ontology and SWRL...... 84

Reviews of Concepts in Knowledge Organization

Birger Hjørland. Classification ...... 97

Knowl. Org. 44(2017)No.2

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 Birger HJØRLAND, Royal School of Library and Information Science, Copenhagen Denmark. E-mail: [email protected] This journal is the organ of the INTERNATIONAL SOCIETY FOR KNOWLEDGE ORGANIZATION (General Secretariat: Amos Michael KLEINEBERG, Humboldt-Universität zu Berlin, Unter den DAVID, Université de Lorraine, 3 place Godefroy de Bouillon, BP Linden 6, D-10099 Berlin. E-mail: [email protected] 3397, 54015 Nancy Cedex, France. E-mail: [email protected]. María J. LÓPEZ-HUERTAS. Universidad de Granada, Facultad de Bib- Editors lioteconomía y Documentación, Campus Universitario de Cartuja, Bib- lioteca del Colegio Máximo de Cartuja, 18071 Granada, Spain. E-mail: Richard P. SMIRAGLIA (Editor-in-Chief), School of Information Stud- [email protected] ies, University of Wisconsin, Milwaukee, Northwest Quad Building B, 2025 E Newport St., Milwaukee, WI 53211 USA. Kathryn LA BARRE, The Graduate School of Library and Information E-mail: [email protected] Science, University of Illinois at Urbana-Champaign, 501 E. Daniel Street, MC-493, Champaign, IL 61820-6211 USA. E-mail: [email protected] Melodie J. FOX (Reviews Editor), Bryant & Stratton College, 310 W. Wisconsin Ave., Ste 500E, Milwaukee, WI 53203 USA. Devika P. MADALLI, Documentation Research and Training Centre E-mail: [email protected]. (DRTC) Indian Statistical Institute (ISI), Bangalore 560 059, India. E-mail: [email protected] Jihee BEAK (Assistant to the Editor-in-Chief), School of Information Studies, University of Wisconsin, Milwaukee, Northwest Quad Building Ia MCILWAINE (Literature Editor), Research Fellow. School of Li- B, 2025 E Newport St., Milwaukee, WI 53211 USA. brary, Archive & Information Studies, University College London, E-mail: [email protected] Gower Street, London WC1E 6BT U.K. E-mail: [email protected] Joshua HENRY (Editorial Assistant), School of Information Studies, Jens-Erik MAI, Royal School of Library and Information Science, Co- University of Wisconsin, Milwaukee, Northwest Quad Building B, 2025 penhagen Denmark. E-mail: [email protected] E Newport St., Milwaukee, WI 53211 USA. Daniel MARTÍNEZ-ÁVILA, Departamento de Ciência da Informação, Universidade Estadual Paulista–UNESP, Av. Hygino Muzzi Filho 737, Editors Emerita 17525-900 Marília SP Brazil. E-mail: [email protected] Hope A. OLSON, School of Information Studies, University of Wis- Elaine MÉNARD, School of Information Studies, McGill University, consin-Milwaukee, Milwaukee, Northwest Quad Building B, 2025 E 3661 Peel Street Suite 303A, Montréal, Quebec, H3A 1X1, Canada. E- Newport St., Milwaukee, WI 53211 USA. E-mail: [email protected] mail: [email protected] Clare BEGHTOL, Faculty of Information Studies, University of To- Widad MUSTAFA el HADI, Université Charles de Gaulle Lille 3, URF ronto, 140 St. George Street, Toronto, Ontario M5S 3G6, Canada. IDIST, Domaine du Pont de Bois, Villeneuve d’Ascq 59653, France. E-mail: [email protected] E-mail: [email protected] Ingetraut DAHLBERG, Am Hirtenberg 13, 64732 Bad Konig,̈ Germa- H. Peter OHLY, Prinzenstr. 179, D-53175 Bonn, Germany. ny. E-mail: [email protected] E-mail: [email protected] Editorial Board K. S. RAGHAVAN, KAnOE (Centre for Knowledge Analytics & Onto- Thomas DOUSA, The University of Chicago Libraries, 1100 E 57th St, logical Engineering), PES Institute of Technology, 100 Feet Ring Road, Chicago, IL 60637 USA. E-mail: [email protected] BSK 3rd Stage, Bangalore 560085, India. E-mail: [email protected]. Jonathan FURNER, Graduate School of Education & Information Studies, University of California, Los Angeles, 300 Young Dr. N, Mail- M. P. SATIJA, Guru Nanak Dev University, School of Library and In- box 951520, Los Angeles, CA 90095-1520, USA. formation Science, Amritsar-143 005, India. E-mail: [email protected] E-mail: [email protected] Aida SLAVIC, UDC Consortium, PO Box 90407, 2509 LK The Hague, Jesús GASCÓN GARCÍA, Facultat de Biblioteconomia i Docu- The Netherlands. E-mail: [email protected] mentació, Universitat de Barcelona, C. Melcior de Palau, 140, 08014 Barcelona, Spain. E-mail: [email protected] Renato R. SOUZA, Applied Mathematics School, Getulio Vargas Foundation, Praia de Botafogo, 190, 3o andar, Rio de Janeiro, RJ, Claudio GNOLI, University of Pavia, Science and Technology Library, 22250-900, Brazil. E-mail: [email protected] via Ferrata 1, I-27100 Pavia, Italy. E-mail: [email protected] Rick SZOSTAK, University of Alberta, Department of Economics, 4 Rebecca GREEN, Senior Editor, Dewey Decimal Classification, Dewey Edmonton, Alberta, Canada, T6G 2H4. E-mail: [email protected] Editorial Office, Library of Congress, Decimal Classification Division , 101 Independence Ave., S.E., Washington, DC 20540-4330, USA. E- Joseph T. TENNIS, The Information School of the University of mail: greenre@oclc.org Washington, Box 352840, Mary Gates Hall Ste 370, Seattle WA 98195- 2840 USA. E-mail: [email protected] José Augusto Chaves GUIMARÃES, Departamento de Ciência da In- formacão, Universidade Estadual Paulista–UNESP, Av. Hygino Muzzi Maja ŽUMER, Faculty of Arts, University of Ljubljana, Askerceva 2, Filho 737, 17525-900 Marília SP Brazil. E-mail: [email protected] Ljubljana 1000 Slovenia. E-mail: [email protected] Knowl. Org. 44(2017)No.2 83 Errata

Errata

Erratum 1 Erratum 2

In volume 43, number 7, page 519 a typo in the formula In volume 44, number 1 (2017) page 55 the citation was was perpetuated from an early version of the manuscript. in error; it should be: We regret the error. Hjørland, Birger. 2016. “Subject (Of Documents).” The correct equation is: Knowledge Organization 44(1): 55-64. 61 references.

, (1)

84 Knowl. Org. 44(2017)No.2 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL†

Xiao-Bo Tang*, Wei Wei**, Guang-Chao Liu*** and Juan Zhu****

*/**/***Center for the Studies of Information Resources, Wuhan University, Wuhan, Hubei, China *, **<[email protected]>, ***[email protected]>, ****Jiujiang University, Jiujiang, Jiangxi, China,

Xiao-Bo Tang is a professor of information science at the Information Management School of Wuhan University in China. He holds a PhD in management science and engineering from Wuhan University and a master’s degree in computer science and technology from Wuhan University of Hydraulic and Electrical Engineering. He is the director of the Information Systems Research Center in Wuhan University, member of the Association for Information Systems, the director of the China Branch of the Association for Information Systems, and the executive director of the Informa- tion Society of Hubei Province. His main research interests include knowledge organization and intelligence analysis.

Wei Wei has been working as a lecturer at Wuhan University since 2007 and she has been a PhD student at the Information Management School of Wuhan University since September 2014. She attained her master’s and bachelor’s degrees in management information systems. Her research interests lie in knowledge organization and intelligence service. She is currently working on knowledge representation and inference applied to ontology in the medical domain as part of her PhD thesis.

Guang-Chao Liu is a doctoral student at the Information Management School of Wuhan University since September 2015. He completed his MS in financial management from the Lubin School of Business, Pace University. He is currently working on text mining in the financial domain at the semantic level. His research interests also include sentiment analysis and knowledge discovery.

Juan Zhu has been a lecturer at Jiujiang University and she has been a doctoral student at the Information Management School of Wuhan University. As part of her PhD thesis, she is currently working towards building knowledge bases to represent the user entity and the item entity for recommender systems. Her research interests also include knowledge organization and intelligence service.

Tang, Xiao-Bo, Wei Wei, Guang-Chao Liu and Juan Zhu. 2017. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL.“ Knowledge Organization 44(2): 84-96. 23 references.

Abstract: Medical insurance fraud is common in many countries’ medical insurance systems and represents a serious threat to the insurance funds and the benefits of patients. In this paper, we present an inference model of medical insurance fraud detection, based on a medical detection domain ontology that incorporates the knowledge base provided by the Medical Terminology, NKIMed, and Chinese Library Classification systems. Through analyzing the behaviors of irregular and fraudulent medical services, we defined the scope of the medical domain ontology relevant to the task and built the ontology about medical sciences and medical service behaviors. The ontology then utilizes Semantic Web Rule Language (SWRL) and Java Expert System Shell (JESS) to detect medical irregularities and mine implicit knowledge. The system can be used to improve the management of medical insurance risks.

Received: 21 August 2016; Revised: 13 January 2017; Accepted 18 January 2017

Keywords: medical insurance frauds, inference model, ontology,

† This study is part of the State National Sciences Foundation Project “Research on Integrated Retrieval and Semantic Analysis Method of Social Media“ (No. 71273194).

Knowl. Org. 44(2017)No.2 85 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

1.0 Introduction As presented by Liu (2014), medical insurance frauds can be classified into internal frauds and external frauds. False medical claims, excessive treatment, forgery of med- Internal frauds are frauds committed by officials or em- ical records and case history, and other methods have ployees of medical insurance or governmental institu- been commonly employed to illegally obtain medical in- tions, and external frauds are those committed by benefi- surance funds from many countries’ medical insurance ciaries or employees of medical service providers. Inter- systems. Those actions can be deemed as medical insur- nal frauds are essentially embezzlement of the insurance ance frauds. According to Zuo (2014), frauds may be fund, and outside the scope of this study, which focuses committed by medical employees, beneficiaries, or by the on external frauds. The main forms of external frauds in- collusion of both. As its medical insurance system conclude: tinues to develop, China has been well on its way to achieving the goal of universal access to health care. – Frauds committed by beneficiaries However, a new problem has emerged that participants – Frauds committed by employees of medical service attempt to take advantage of the system’s loopholes. In providers order to identify medical insurance frauds, we propose an – Collusion between beneficiaries and employees of inference model of medical insurance fraud detection providers based on ontology and SWRL technologies. As Cristiane (2016) explained, ontology is an effective method for rep- In principle, the medical insurance system is created to resenting the knowledge of a given field. It can provide mitigate the beneficiaries’ financial damage caused by an extremely powerful, meta-level resource that uses in- health risks. But this does not prevent some beneficiaries ference to analyze explicit knowledge in the ontology. from attempting to exploit the system. Common behav- The model presented in this study is aimed at detecting iors include imposture by non-participants of medical in- medical irregularities by mining implicit knowledge based surance, providing non-factual information in order to on the characteristics of medical insurance. It can be used receive excessive or unnecessary services, and forgery of to improve the quality of medical risk management and medical records for indemnification. Providers of ser- prediction. vices consist of designated pharmacies and designated The remainder of this article is arranged as follows: health care institutions. As shown by Zhang (2014), section 2 is a literature review that summarizes back- frauds committed by employees of the provider account ground knowledge related to medical insurance frauds, for the largest proportion of medical frauds. These ontology, OWL, SWRL and JESS. Section 3 describes the mainly take the forms of admitting patients who do not general design goals and framework of the system. Sec- need hospitalized treatment, excessive services including tion 4 describes the construction of the ontology, and inspections, treatments, and drug usage, and illegal trans- section 5 describes the working of SWRL rules and JESS actions of documents and prescriptions related to medi- inference engine. The method is tested in section 6 using cal insurance. It is also possible for providers and benefi- information from 200 patient cases to detect medical ir- ciaries to conspire in medical frauds. These include issu- regularities from their medical treatment records and ing fictitious items that require insurance payment, re- validate results. Section 7 consists of the concluding re- porting diseases not applicable for insurance (e.g., traffic marks and problems for future research. accidents, fights, etc.) as applicable, and forgery of medical information (Li 2006). 2.0 Literature review 2.2 Ontology 2.1 Medical insurance frauds Ontology is a formal description of concepts and their Medical insurance frauds include any behavior aimed at relations in a specific domain (David et al. 2009), consist- gaining illegal access to medical insurance funds by using ing of concepts, attributes, and instances (Chakkrit and forged information or concealing truth. In China, the Michael 2009). Ontologies are built to share and reuse primary cause of medical insurance frauds is the “third- knowledge using artificial intelligence and machine lan- party payment” system. Specifically, the payment for the guage and to facilitate communications between humans providers of medical services is not directly paid by the and computer systems. They have been widely utilized for beneficiaries of said services, but entirely by third-party information retrieval and knowledge organization. Due to medical insurance institutions. The frauds are motivated the complexity and diversity of medical knowledge, by the financial gain of medical service providers and ben- building the domain ontology of medical sciences would eficiaries. enable us to establish a knowledge set with unambiguous 86 Knowl. Org. 44(2017)No.2 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL specifications and semantic features. This set can describe 2.5 JESS medical concepts, their relationships, and the general principles of medical science. Researchers have proposed more specialized inference Much work has been done on ontologies for medical engines such as Racer, FaCT and Pellet. While efficient sciences. The Unified Medical Language System (UMLS) and easy to use, they are not very expandable or com- (http://www.nlm.nih.gov/research/umls/) was devel- patible with other languages. This team chooses JESS for oped by the National Library of Medicine of the United its efficiency, flexibility and excellent in porting and em- States to solve the problem of different medical sources bedding. Developed by Friedman-Hill et al., JESS (Java having different expressions for the same concept, and is Expert System Shell) is a rule engine and scripting envi- capable of providing semantic unification of medical ronment written entirely in Java (Martin et al. 2005). The concepts. The Open Biomedical Ontology (OBO) was core JESS language is compatible with CLIPS (C Lan- built by Barry et al. (2007) from the National Institute of guage Integrated Production System), which uses the Health of the United States, and consists of multiple sets Rete algorithm to process JESS rules. Rete is a high- of knowledge related to biology and medicine, including efficiency algorithm used to solve many matching prob- the gene ontology (GO), plant ontology (PO), cell type lems, particularly known for its speed in inference opera- ontology (CT), sequence ontology (SO) and more. In tions. JESS has inherited the advantages of CLIPS, and China, researchers have developed systems including added some new features, such as forward inference, re- NKIMed (Zhou 2003), the Traditional Chinese Medical verse inference, running memory check, and the ability Language System (Zeng and Wang 2006), and the Chi- of directly calling and operating a Java class library. JESS nese medical information semantic indexing system and also supports the traditional “if … then …” grammar semantic retrieval model (Li and Pang 2003). In general, structure. the theories on medical ontologies have become increasingly more mature, and the practical application of medi- 3.0 Research framework cal ontology is also attracting widespread attention. 3.1 Targets of detection 2.3 OWL This study is aimed at developing a model for the expres- According to Mercedes (2008), OWL (Web Ontology Lan- sion and inference of behaviors related to medical insur- guage) is a framework proposed by the World Wide Web ance, in order to detect and prevent irregularities. As pre- Consortium (W3C), and a recommended language for se- sented by Li (2015), medications, inspections, and treat- mantic web-based ontology. It is an extension of RDF ments are important components of medical services, (Resource Description Framework) that builds on XML- and also the main source of frauds. The model proposed based RDF syntax structure and is designed for description by this study targets the behaviors detailed in Table 1. logic. OWL is fully compatible with RDF, complements RDF’s deficiency in describing relationships, and is better Irregular Prescriptions from e.g., prescription suited for expression and inference. Based on purposes, medications service providers in- contains multiple clude excessive drugs with simi- three different levels of OWL can be used: OWL Full, amounts of drugs or lar or the same OWL DL and OWL Lite (http://www.w3.org/TR/owl- drugs unrelated to effects features/). the disease Excessive Services are unneces- e.g., unnecessary 2.4 SWRL services sary or beyond the inspections were regulations related to performed; hos- basic medical insur- pitalization for SWRL (Semantic Web Rule Language) was proposed to ance trivial diseases complement OWL’s deficiency in ontology inference. As Multiple One large prescrip- e.g., several un- presented by Jeff (2005), the current W3C standard pro- prescriptions tion is divided into necessary pre- vides a semantic description of rules allowing users to several smaller ones scriptions are build a rules-based ontology, and achieve semantic infer- made for the ence for the ontological knowledge base. By describing treatment of one disease the relationships between rules, SWRL can improve the ontology’s capability for semantic expression. Its canoni- Table 1. Targets of detection. cal format is antecedent → consequent. “Antecedent” and “consequent” are disjunctions of atoms in ontology, and All of the irregularities take place as patients receive medi- atoms are classes or attributes from the ontology. cal services from the medical service institutions. For this Knowl. Org. 44(2017)No.2 87 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

Figure 1. Framework of medical insurance fraud detection.

reason, the detection of the inference model should pri- ontology into an acceptable format for operating on the marily be performed on the medical records, prescriptions, medical instances. Because SWRL cannot be operated on and inspection documents issued by health service provid- directly, the SWRL rules need to be converted through ers. It is relatively difficult to discover medical insurance- XSLT (Extensible Stylesheet Language) into a format that related misconducts by simply matching strings or calculat- can be executed by JESS. Additionally, Pellet can also de- ing numerical differences. This model instead approaches tect inconsistencies and conflicting knowledge in the on- the problem by analyzing the patients’ medical activities us- tology (Rung et al. 2012). ing the knowledge representation of medical behaviors and inferring the relationships intrinsic to the knowledge. 4. JESS is used for the inference process, to identify which classes, axioms or instances are applicable to the 3.2 Model framework current medical service behavior. OWL and SWRL are converted into a format acceptable for JESS during the Figure 1 shows the inference model for fraud detection. process; afterwards the results of inference are written in OWL for updating the medical detection ontology. 1. The medical fraud detection ontology normalizes the knowledge of medical service behaviors and expert 4.0 Construction of medical detection domain knowledge from the medical domain. The preprocessed ontology data of instances are added to the ontology through a medical information system to create a knowledge base, In this study, we construct a medical fraud detection on- as shown in Figure 2. tology mainly used for detecting illegal behaviors related to the medical insurance industry by beneficiaries and 2. The inference rules of medical fraud detection are de- service providers. The ontology utilizes Medical Termi- scribed using SWRL. The SWRL rules are built with ex- nology (Medical Term Validation Committee 1998), isting classes, attributes and instances from the ontology. NKIMed, and Chinese Library Classification (Editorial The system will warn users of irregularities found in the board of the National Library 2010) to define the termi- medical information, such as repeated prescriptions of nologies and concepts in the field of medical detection. the same drug, and excessive inspections. Terminologies are the basic elements of the ontology, used to describe the knowledge representations of the 3. The Pellet inference engine is used to convert the concepts. Table 2 shows a selection of important terms knowledge classes and concepts of the medical detection used in this study. 88 Knowl. Org. 44(2017)No.2 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

Figure 2. The development of domain ontology.

Medical in- doctors, patients, diseases, drugs, symp- The ontology is constructed using Protégé, a Java-based, formation toms, inspections, treatments open source graphical application, developed by Stanford Diseases endocrine disease, respiratory disease, University Information Center for accessing, creating and cardiovascular diseases, digestive diseases, viral diseases, dermatology diseases, oph- maintaining ontologies. Protégé uses RDF to describe the thalmology disease, nervous diseases, relationships between web pages and other resources, and bacterial infections and fungal diseases, to support description of ontologies’ main components, otorhinolaryngology diseases, cancer classes and attributes. The main components of Protégé Drugs chemical medicine, Chinese patent medicine, Chinese herbal medicine, endocrine are OWL classes, properties, and individuals and so on. system medicine, oral medicine, respira- Classes are the semantic representations of concepts in tory medicine, cardiovascular medicine, knowledge. Properties describe the relationships of class digestive medicine, vitamin drugs, co- linked to the basic data types, and the relationships can be deine phosphate, aspirin, Proglumide connected to other classes or instances. Individuals repre- Inspections general clinical test, medical imaging, ul- trasonic inspection, blood pressure check, sent objects in a domain that can be considered as in- blood inspection, bone marrow inspec- stances of classes. Property here is a bilateral relationship tion, urine inspection, CT scan, B-scan between two individuals, or can be considered as a bridge ultrasonography between the two individuals, making it different from the Treatments surgical treatment, medication, chemo- therapy, physic therapeutics, cinesiatrics, attributes in an object-oriented programming language. acupressure treatment, acupuncture, die- Protégé has good support for ontology languages includ- tary therapy, blood transfusion, biother- ing RDF, OWL, and Schema XML. Being Java-based, it apy ensures the expandability of the ontology’s application Symptoms headache, astriction, haemorrhage, gasp, environment, and provides fast compiler for ontology hematochezia, sharp pain, back pain, lumbago, vertigo, macula, arthronalgia, construction. frequent micturition, fever, claudication, Figure 3 shows the medical detection ontology built in odontalgia Protégé, with the classes and their hierarchical structure Irregularities split prescription, excessive services, ir- to the left, and the specified constraints to the right. regular medication, similar drug efficacy, excessive prescriptions We adopted the knowledge engineering method (Nata- lya and Deborah 2006) to construct the medical detection Table 2. Some important terms in the medical insurance fraud domain ontology. Knowledge engineering is an approach detection domain ontology. Knowl. Org. 44(2017)No.2 89 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

Figure 3. Medical detection domain ontology created in Protégé. to ontology development for Protégé proposed by Nata- identified. The design of the structure was done by con- lya et al. at Stanford University. They believed that there sulting medical experts and standards including Chinese Li- is no absolutely correct way to model a domain, all solu- brary Classification and NKIMed. A total of nine main tions must be adapted to practical application, and the classes are used: medical information, doctors, patients, process of ontology development is one of continued it- symptoms, inspections, diseases, treatments, drugs and ir- eration. The method consists of the process of deter- regularities. Each concept is then classified and arranged mining the ontology’s domain and scope, enumerating es- into a hierarchical structure with super-classes and sub- sential concepts, defining classes and their hierarchical classes (Subhashis and Sayon 2016). Figure 4 shows the structure, defining attributes, creating instances in the on- medical detection domain ontology model. tology, and evaluating the ontology. The following discussion describes the construction of the medical detection 4.2 Defining attributes ontology in this study. The definition of classes and their hierarchy alone cannot 4.1 Defining classes represent all knowledge information in a field, as their internal structures must also be described by defining at- Classes are the core of ontology, providing an abstract de- tributes. Object attributes and datatype attributes are two scription of the physical objects in the domain. The estab- important attribute types in an OWL ontology. Object at- lishment of the medical insurance fraud detection domain tributes represent the mutual relationships between two ontology is based on current medical domain knowledge, classes, and a class’s attributes are linked to instances allowing us to define the general concepts first, and then through its domain and range. Datatype attributes are narrow down the definitions to suit our need. The classes used to designate a class’s unique attributes, and represent and hierarchical structure of the ontology should provide a the relationship between instances and data. In this study, comprehensive coverage of topics related to behaviors in the important attributes of the medical detection domain medical services, before its issues and constraints can be ontology include the purpose of detection, treatment 90 Knowl. Org. 44(2017)No.2 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

Figure 4. The model of medical detection domain ontology. measures, and related diseases, which are object attrib- 5.0 The inference of medical detection domain utes, and admission IDs, times of admission, and patient ontology names, which are datatype attributes. Inference means obtaining the implicit knowledge from a 4.3 Adding instances given set of knowledge, and ontology inference involves extracting the implicit knowledge from explicit knowl- A class can contain multiple instances, while one instance edge defined in ontology. Ontology inference can serve can belong to one or more classes. In this study, instances functions including semantic-based query, knowledge in- were added according to the need of the experiment. ference, and ontology checking. It is based on the idea of Medical information instances were created for testing description logic. Description logic is a logic-based know- the system’s inference rules. ledge representation, and a decidable subset of first-order logic, that can describe the provability of a given state- 4.4 Integrity and consistency test ment within limited time. It consists of concepts, attributes or roles, and individuals. The basic problems of de- A robust ontology is fundamental for the proper execu- scription logic inference include the inclusion relation of tion of the inference model. The Pellet inference engine concepts, the satisfiability problem, and entity detection can detect conflicts in the ontology, and was used to test (Shi and Sun 2006). the integrity and consistency of the ontology, as shown in Adela et al. (2009) show that much work has been Figure 5. done on inference engines for description logic-based in- The construction of ontology is by necessity a com- ference. FaCT, Racer and Pellet are several inference en- plex process that requires cooperation between experts gines based on OWL. Figure 6 depicts the workflow of and knowledge engineers. Iterative testing and improve- inference in this study. The main components of the in- ment are needed to gradually add details and optimize the ference engine are the parser, the tuples, and the rules. ontology. The ontology constructed in this study is only First, the engine reads the ontology in OWL format. The used as an experimental model to verify the validity of OWL parser then converts the ontology from OWL into the inference system. the tuples format of < subject, verb, object >. Rules are restrictions that focus the ontology on a particular domain. Finally, the ontology of classified OWL is output by the description logic engine. Knowl. Org. 44(2017)No.2 91 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

Figure 5. Consistency test in Pellet.

Figure 6. The workflow of inference engine.

Our study uses SWRL format for the inference rules of Suppose an efficacy relationship has drugs as its domain, medical fraud detection, which have XML-based syntax. and diseases as its range, with the knowledge that two dif- SWRL rules can supplement the OWL ontology with ro- ferent drugs x1 and x2 can both cure disease y1, therefore bust semantic expressions, creating inference rules based x1 and x2 have the same efficacy, described as: on the instances and attributes in the ontology. SWRL rules can directly reference the ontology’s classes, attrib- hasEfficacyOf(?x1,?y1)∧hasEfficacyOf(?x2,?y1)∧ utes and relationships. For example, the following rela- differentForm(x1,x2) → hasSimilarEfficacy(x1, x2) tionships have been defined in the ontology: SWRL allows us easy editing of inference rules using in-

hasEfficacyOf(x1, y1) stances and attributes in the ontology through the SWRL

hasEfficacyOf(x2, y1) Tab plug-in in Protégé. However, as the inference process 92 Knowl. Org. 44(2017)No.2 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

Figure 7. Workflow of JESS inference format conversion. cannot be performed directly from Protégé, it is neces- record contains the patient’s conditions (the patient’s basic sary to use the ontology knowledge facts as the basis, and information as well as some physiological parameters) and infer the rules through the rules engine. the prescriptions given by the doctor, and the final infer- For this study we chose JESS (Java Expert System ence results will be compared with the experts from Medi- Shell) as the inference engine due to its good portability, cal Insurance Regulatory. embeddedness and high efficiency. While JESS cannot di- By selecting the SWRL Tab plug-in, the SWRL rules rectly read ontologies from the OWL format or rules can be edited in Protégé on the OWL platform. This study from the SWRL format, the inference can be done after uses JESS as a rules engine to embed the Protégé platform performing format conversions. Figure 7 shows the for- to perform rules inference. JESS software consists of the mat conversion and inference process for JESS. First, the rules base, the fact base and the execution engine (Jing and ontology is converted to JESS by Pellet inference engine. Elena 2004). Figure 8 shows SWRL’s operation interface. Second, the SWRL rules are converted to JESS by XSLT The two rectangular boxes with dotted lines are the SWRL (Stylesheet Language Extensible), and the inference re- rules editor and the JESS operation interface. The rules are sults are written to the OWL to update the ontology. written in the editor, and the JESS operation interface provides the triggers for calling the rules engine. 6.0 Experiment and discussion We developed four SWRL rules for medical fraud detection. The rules can be expressed as the following. 6.1 Constructing the medical insurance fraud detection system RULE 1: If drug x1 can cure disease y1, and drug x2 can also We used Protégé to build the medical insurance fraud de- cure disease y1, then they have similar efficacy, and a tection ontology, with the hierarchical structure shown in case of irregular medication is detected. the previous sections. Figure 3 depicts the Protégé OWL ontology editor. The root node of the ontology is “owl: Medical_information(?I) ∧ Drugs(?x1) ∧ thing,” and its child nodes consist of the nine subclasses: UseDrugs(?I, ?x1) ∧ Drugs(?x2) ∧UseDrugs(?I, ?x2)∧ medical information, doctors, patients, symptoms, inspec- Symptoms(?y1)∧hasEfficacyOf(?x1, ?y1)∧ tions, diseases, treatments, drugs and irregularities. Two hasEfficacyOf(?x2, ?y1) →hasIrregularities(?I,SimilarEfficacy) hundred medical treatment records were established to ver- ∧IrregularitiesOrNot(?I, true) ify the effectiveness of the rules inference. Each medical Knowl. Org. 44(2017)No.2 93 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

Figure 8. The joint interface of SWRL rule editing and JESS operation.

RULE 2: then a case of multiple split prescriptions for one If the actual dosage of a drug exceeds its proper dos- beneficiary is detected. age, then a case of excessive prescription is detected.

Medical_information(?i1) ∧Medical_information(?i2)

Medical_information(?i) ∧ Drugs(?x) ∧UseDrugs(?i, ?x) ∧ ∧ Patients(?p1) ∧isPatientOf(?i1, ?p1) ∧ Patients(?p2)

Diseases(?d) ∧hasEfficacyOf (?x, ?d) ∧ ∧isPatientOf(?i2, ?p2) ∧PatientID(?p1, ?id)

isDiseaseOf(?i, ?d) ∧MedicalDosage(?i, ?y) ∧ ∧PatientID(?p2, ?id) ∧ Diseases(?d) ∧isDiseaseOf(?i1, ?d)

Dosage(?x, ?z) ∧swrlb:lessThan (?z, ?y) →hasIrregularities(?i, ∧isDiseaseOf(?i2, ?d) →hasIrregularities(?i1, ExcessivePescriptions) ∧IrregularitiesOrNot(?i, true) MultiplePrescriptions) ∧hasIrregularities

(?i2,MultiplePrescriptions) ∧IrregularitiesOrNot(?i1, true)

RULE 3: ∧IrregularitiesOrNot (?i2, true) If the medical information entry i contains separate inspections for clinical responses and disease determi- Through the implementation of the “J” button, the JESS nation, then a case of excessive inspection is detected. engine is triggered. Figure 9 shows the integration of OWL ontology and SWRL, and the inference results Medical_information(?i) ∧Test_items(?e) ∧hasBodytest(?i, from JESS. The results show that the process had se- ?e) ∧isTestOf(?e, ?z) ∧Symptoms(?y) ∧ lected 4 rules, 64 classes and 59 individuals. hasClinicalResponse(?i, ?y) ∧ Diseases(?d) ∧isDiseaseOf(?i, It can be seen that nine axioms have been added, ?d) ∧differentFrom(?z, ?y) ∧differentFrom(?z, ?d) → which are the results of JESS inference. As shown in hasIrregularities(?i,ExcessiveInspections) ∧ Figure 10, the instance “Medical_information_01“ has its IrregularitiesOrNot(?i, true) attribute “hasIrregularities” assigned the value “true,” with the attribute “SimilarEfficacy“ attached. This shows RULE 4: the validity of the inference. If the medical information entries i1 and i2 for pa- The mechanism of medical data detection is based on tients p1 and p2 have the same citizen’s ID number, medical fraud detection domain ontology. The medical 94 Knowl. Org. 44(2017)No.2 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

Figure 9. The inference result of OWL+SWRL→Jess.

Figure 10. The inference result of medical fraud detection ontology.

treatment data can be processed based on the rules of formance. The test was performed on two hundred med- detection. Finally, the obtained explicit knowledge is pre- ical records to check the system’s precision and recall. served in the knowledge base, which can be parsed and Precision refers to the percentage of correctly detected extracted for inquiries. (i.e. true) irregular records among the detected records, and recall refers to the percentage of correctly detected 6.2 Evaluating the medical insurance fraud records among all irregular records. detection system In Table 3, the True Positive rate (TP) represents the percentage of cases where the experts agreed with the We applied the system to the sample data, and had ex- detected results. The False Negative rate represents cases perts on medical insurance regulation evaluate its per- where the experts disagreed that the detected results are Knowl. Org. 44(2017)No.2 95 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

Parameter Definition model, and use the ontology technology to detect fraudu- True Positive rate The system detects and the expert lent behaviors. This model is an experimental model based (TP) agree on knowledge of medical domain that can discover irregu- False Negative rate The system detects but the expert larities in the rendering of medical services. It provides a (FN) does not agree new attempt at the detection of medical insurance fraud False Positive rate The system does not detect but behaviors. (FP) the expert detects Irregularities in medical services involve a wide range of Table 3. Evaluation of the medical insurance fraud detection knowledge, including the efficacy, pharmacology, indica- system. tions of drugs, and clinical representations of diseases, and much more. Obtaining such knowledge requires tight co- irregularities. The False Positive rate (FN) represents cases operation between experts and knowledge engineers. This where the experts believed the system had failed to detect study mainly focuses on the effective representation of irregularities. The precision rate is determined by dividing knowledge in the medical domain and the technical im- TP with the sum of TP and FN, shown by Equation (1). plementation of ontology-based inference. Our future Equation (2) shows the recall rate, determined by dividing work can focus on further improvements on the ontologi- TP with the sum of TP and FP. cal knowledge base of medical insurance fraud detection,

designing more rules to cover additional illegal behaviors Precision=TP/ (TP + FN) (1) and comparing the results with other inference engines and Recall = TP / (TP+ FP) (2) methods, and extending the model to adopt it to big data

medical platforms. Table 4 represents the detected results of the system and the experts. The system achieved a precision rate of References 100% and a recall rate of 92.86%. From two hundred medical information instances, it detected thirteen entries Baorto, David, Li Li and James J. Cimino. 2009. “Practical with irregularities, which are all consistent with the expert Experience with the Maintenance and Auditing of a findings. The experts also discovered one additional re- Large Medical Ontology.“ Journal of Biomedical Informat- cord with irregular behavior; this is due to that the rules ics 42: 494-503. did not account for the misconduct involved in this re- Casteleiro, Mercedes Argüello and Jose Julio Des Diz. cord, which can be remedied by further refining the rules. 2008. “Clinical Practice Guidelines: A Case Study of The consistency proves the validity of the system. Combining OWL-S, OWL, and SWRL.“ Knowledge-Based

Systems 21: 247–255. System Experts Chen, Rung-Ching, Yun-Hou Huang, Cho-Tsan Bau and Similar Efficacy 7 7 Shyi-Ming Chen. 2012. “A Recommendation System Excessive Prescriptions 3 3 Based on Domain Ontology and SWRL for Anti- Excessive Inspections 1 1 diabetic Drugs Selection.“ Expert Systems with Applications Multiple Prescriptions 2 2 39: 3995-4006. Other Irregularities 0 1 Das, Subhashis and Sayon Roy. 2016. “Faceted Ontologi- Table 4. The evaluation results of the system by the experts. cal Model for Brain Tumour Study.“ Knowledge Organi- zation 43: 3-12. 7.0 Conclusions and prospective Editorial board of the National Library. 2010. China Li- brary Classification. Beijing: National Library Press. The medical insurance system is a foundational institution OCLC has Chinese Library Classification = Zhong guo tu of national welfare that ensures citizens’ access to medical shu guan fen lei fa shi yong shou ce / $c Zhong guo tu services. The risks of frauds in the field have grown with shu guan fen lei fa bian ji wei yuan hui bian. Di 1 ban. the development of the industry, and their prevention and Bei jing: Shu mu wen xian chu ban she, 1999. control are crucial to saving people’s lives. Thus the detec- Laua, Adela, Eric Tsuib, and W. B. Lee. 2009. “An Ontol- tion of illicit conduct, the discovery of explicit relation- ogy-Based Similarity Measurement for Problem-Based ships and the prevention of irregularities in the medical Case Reasoning.“ Expert Systems with Applications 36: service process have become important topics of research 6574-79. by medical insurance administration and related research- Lian, Shi and Sun Jigui. 2006. “Description Logic Sur- ers. Our study presents an analysis of frauds in medical in- vey.“ Computer Science 33: 194-97. surance, from the perspective of medical service behaviors. Medical Term Validation Committee. 1998. Medical Termi- We propose a medical insurance fraud detection inference nology. Beijing: Science Press. 96 Knowl. Org. 44(2017)No.2 X. Tang, W. Wei, G. Liu and J. Zhu. “An Inference Model of Medical Insurance Fraud Detection: Based on Ontology and SWRL

Mei, Jing and Elena Paslaru Bontas. 2004. “Reasoning Goldberg, Karen Eilbeck, Amelia Ireland, Christopher Paradigms for OWL Ontologies.“ Technical Report B- J. Mungall, The OBI Consortium, Neocles Leontis, 04-12. Berlin: AG Netzbasierte Informationssysteme, Philippe Rocca-Serra, Alan Ruttenberg, Susanna- 1-23. Assunta Sansone, Richard H Scheuermann, Nigam Netto, Cristiane Mendes, Gercina Ângela Borém de Oli- Shah, Patricia L Whetzel and Suzanna Lewis. 2007. veira Lima and Ivo Pierozzi Júnior. 2016. “An Applica- “The OBO Foundry: Coordinated Evolution of On- tion of Facet Analysis Theory and Concept Maps for tologies to Support Biomedical Data Integration.“ Na- Faceted Search in a Domain Ontology: Preliminary ture Biotechnology 25: 1251-55. Studies.“ Knowledge Organization 43: 254-64. Snae, Chakkrit and Michael Brueckner. 2009. “Personal Ni, Zuo and Chu Lifeng. 2014. “Analysis of Monitoring Health Assistance Service Expert System (PHASES).“ Health Care Fraud based on the Data Analysis Tech- International Journal of Biological and Medical Sciences 26: nology.“ Insurance Practice and Exploration 11: 32-40. 157-60. Noy, Natalya F. and Deborah L. McGuinness. 2006. “On- Xiaobin, Zhou and Cao Cungen. 2003. “Medical Knowl- tology Development 101: A Guide to Creating Your edge Acquisition: An Ontology-Based Approach.“ First Ontology.“ http://protégé.stanford.edu/publica Computer Science 10: 40-45. tions/ontology_development/ontology101-noymc Xin, Li. 2006. “Analysis of Potentially Fraudulent Behav- guinness.html ior of Basic Medical Insurance.“ Journal of Hubei Uni- O’Connor, Martin Joseph, Holger Knublauch, Samson versity of economics 3 no. 9: 40-41. Tu and Mark Alan Musen. 2005. “Writing Rules for Xinmin, Zhang. 2014. “Study on the Legal Liability Sys- the Semantic Web using SWRL and JESS.“ Proceedings tem of Social Medical Insurance Fraud.“ Journal of of the 8th International Protégé Conference, July 18-21, 2005, Southwest University for Nationalities 35: 90-95. Madrid, Spain. Protege with Rules Workshop, 1-7. Yazi, Li and You Bin. 2015. “Analysis of Characteristics Pan, Jeff Z., Giorgos Stoilos, Giorgos Stamou, Vassilis of Medical Insurance Fraud.“ China Social Security 2015 Tzouvaras and Ian Horrocks. 2005. “f-SWRL: A Fuzzy no. 2: 76-79. Extension of SWRL.“ In Artificial Neural Networks: Yi, Li and Pang Jingan. 2003. “Research on Semantic In- Formal Models and Their Applications- ICANN 2005: 15th dexing System and Semantic Retrieval Model for Chi- International Conference, Warsaw, Poland, September 11-15, nese Medical Information based on Multilayer Concep- 2005, Proceedings, Part II, ed. Wlodzislaw Duch, Erkki tual Semantic Network Structure.“ Journal of the China Oja and Slawomir Zadrozny. Berlin: Springer, 829-34. Society for Scientific & Technical Information 22: 403-11. Ruyi, Liu. 2014. “Research and Application of Medical Zhao, Zeng and Wang Xiaoping. 2006. “UMLS and the Violation Reasoning Detection Model Based on On- Establishment of the Traditional Chinese Medical tology.“ Masters thesis, Jiangsu University. Language System.“ Chinese Journal of Medical Library and Smith, Barry, Michael Ashburner, Cornelius Rosse, Jona- Information Science 15 no. 3: 1-3. than Bard, William Bug, Werner Ceusters, Louis J

Knowl. Org. 44(2017)No.2 97 Reviews of Concepts in Knowledge Organization

Reviews of Concepts in Knowledge Organization

Series Editor: Birger Hjørland

* † Classification

Birger Hjørland

University of Copenhagen, Royal School of Library and Information Science, Njalsgade 76, DK-2300 Copenhagen S, Denmark,

Birger Hjørland holds an MA in psychology and PhD in library and information science. He is Professor in knowledge organization at the Royal School of Library and Information Science in Copenhagen since 2001 and was Professor at the University College in Borås 2000-2001. He was research librarian at the Royal Library in Copenhagen 1978-1990, and taught information science at the Department of Mathematical and Applied Linguistics at the University of Copenhagen 1983-1986. He is chair of ISKO”s Scientific Advisory Council and a member of the editorial boards of Knowledge Organization, Journal of the Association for Information Science and Technology and Journal of Documentation. His H-index is 41 in Google Scholar and 23 in Web of Science.

Hjørland, Birger. 2017. “Classification.” Knowledge Organization 44(2): 97-128. 166 references.

Abstract: This article presents and discusses definitions of the term “classification” and the related concepts “Concept/conceptualization,” “categorization,” “ordering,” “taxonomy” and “typology.” It further presents and discusses theories of classification including the influences of Aristotle and Wittgenstein. It presents different views on forming classes, including logical division, numerical taxonomy, historical classification, hermeneutical and pragmatic/critical views. Finally, issues related to artificial versus natural classification and taxonomic monism versus taxonomic pluralism are briefly presented and discussed.

Received: February 4, 2017; Accepted: February 7, 2017

Keywords: classification, theory, knowledge, concepts

* Derived from the article of the same title in the ISKO Encyclopedia of Knowledge Organization, version 1.1; published 2017-02-03. Article category: core concepts.

† I am very grateful to Fulvio Mazzocchi, who served as the editor of this article. He and the two anonymous referees provided detailed, knowledgeable, careful, and fruitful suggestions for improving the original manuscript. Also thanks to Daniel Parrochia for positive feedback on the manuscript.

1.0 Introduction organization systems (KOSs). These activities and systems are based on more fundamental conceptions and This article is about classification as a basic term in an in- theories of classifications that are presented in this arti- terdisciplinary perspective. Classification is a fundamental cle. concept and activity in knowledge organization, but it is The ISKO Encyclopedia of Knowledge Organization also an important concept in many other fields, including (IEKO) plans to cover a very broad spectrum of articles biology and philosophy. In knowledge organization and related to classification besides the present one. We already library and information science (LIS), it is mostly about have an article about logical division (http://www.isko.org/ classifying documents, document representations, and cyclo/logical_division), and further articles are planned concepts (e.g., in thesauri), and library classification sys- about, for example, library classification, automatic classifi- tems and ontologies are well-known kinds of knowledge cation, numerical taxonomy, classification of the sciences, 98 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization classification in specific domains (including biology, phys- late them in some system according to some princi- ics and chemistry), and much more. ple or conception, purpose or interest …. These This article covers a very complex concept and is three processes, classing, forming classes, and arrang- therefore highly compressed and abbreviated. In particu- ing classes, are so implicated that it is not easy to lar, most of the theories mentioned in section 4 deserve separate them in thought or in terminology; yet we to be enlarged on in independent articles that it is hoped propose here that this should be done as conductive will be forthcoming at some point in the future. to precision in this study. There is an important distinction between assigning a thing, or things, to some 2.0 The meaning of the word “classification” class or classes, and arranging classes in some order or system .… Among the many contributors to the definition of “classification,” two (Frederik Suppe and Henry E. Bliss) are 3) A classification is a series or system of classes ar- here selected as outstanding. Frederick Suppe distin- ranged in some order according to some principles guished two senses of classification: a broad and a nar- or conception, purpose or interest, or some combi- row meaning. He called the broad meaning “conceptual nation of such. classification (1989, 292 emphasis original): There are many more definitions of classification than Classification is intrinsic to the use of language, the ones given above. For a chronological sample of hence to most if not all communication. Whenever definitions of classification, see the Appendix. we use nominative phrases we are classifying the des- The objects we classify may be physical objects, per- ignated subject as being importantly similar to other sons, processes, ideas, concepts, words, etc. Some of entities bearing the same designation; that is, we clas- these entities, such as concepts, may be both the elements sify them together. Similarly the use of predicative classified and a result of a (new) classification. phrases classifies actions or properties as being of a particular kind. We call this conceptual classification, Results of classification What may be classified: since it refers to the classification involved in con- may be termed: ceptualizing our experiences and surroundings. Concepts Categories Documents Clades Classification in the narrower meaning Suppe called “sys- Elements Classes tematic classification” (292): “A second, narrower sense Entities Concepts of classification is the systematic classification involved in Ideas (including fictional Genera the design and utilization of taxonomic schemes such as ideas) the biological classification of animals and plants by ge- Individuals Groups nus and species.” Henry E. Bliss (1929, 142) also considered the senses Items Kinds of the word “classification” and wrote: “this term, like Objects Sets other English derivatives ending in ion, is ambiguously Phenomena Sorts used both in the predicative and in the substantive sense, Processes Species now for the action and now for the act, sometimes for Sciences Taxa the process and sometimes for the product.” In order to Things Etc. remove this ambiguity, he suggested three definitions Etc. proceeding from “class” as a substantive (142-43): Table 1. Selected terms used about the units classified and the 1) The verb to class denotes likening, referring, or resulting groups. assigning a thing to some class, or several things to their respective classes, as may be requisite or rele- The objects to be classified have attributes with values. vant to interest involved. This verb is used not only Attributes may, for example, be color or weight. Values transitively, but sometimes intransitively. Thus it may be red or heavy. Classifications are made by consid- may be said that olive oil classes as a luxury. ering different attributes and their values.1 In conclusion, “classification” is a term used both about 2) The verb classify means primarily to make, or con- the process to classify (which is a kind of discriminative ceive, a class, or classes, from a plurality of things, and practice; see Schmidt and Wagner 2004, 392) and about the secondary to arrange classes in some order or to re- resulting set of classes, as well as the assignment of ele- Knowl. Org. 44(2017)No.2 99 Reviews of Concepts in Knowledge Organization ments to pre-established classes. The wide meaning of truth based upon unitary descriptions of concepts, classi- “classification” is the process of distinguishing and distrib- fication systems represent only particular points of view. uting kinds of “things” into different groups. All narrower She hereby supports the argument of Hjørland and meanings of “classification” are based on the wide defini- Albrechtsen (1999) and Beghtol (2003) that classification tion but add some extra requirements or restrictions put to research must be situated within specific contexts and the the classification process and the resulting classification domains in which the classification systems are designed system—for example, the requirement that a classification to function, as well as Mai’s (2004, 41) claim that “Any should use only one criterion of division at a time, that classification is relative in the sense that no classification classes should be mutually exclusive, and jointly exhaustive, can be argued to be a representation of the true structure are requirements demanded by some specific theories of of knowledge … a classification is merely one particular classification, but not requirements that are common for all explanation of the relationships in a given field that satis- kinds of classification as here defined. fies a group of people at a certain point in time.” Frické (2012, 33), however, is opposed to considering 3. Related terms concepts as mental constructs, and writes that the word “concepts” “amounts roughly to “general notion” or 3.1 Concept/conceptualization “general idea” or even “meaning.” Many describe concepts as being mental or mental constructions; however, “Concept” has formerly been defined the following way we regard them as abstractions or abstract objects (in the (Hjørland 2009, 1522-23): standard Fregean third realm).” In spite of this close connection between classification Concepts are dynamically constructed and collec- and concepts, the discourses on concepts and the dis- tively negotiated meanings that classify the world courses about classification seem mostly to be separated according to interests and theories. Concepts and in the literature. their development cannot be understood in isolation from the interests and theories that motivated 3.2 Categorization their construction, and, in general, we should expect competing conceptions and concepts to be at Elin K. Jacob found that classification and categorization play in all domains at all times. are different processes (2004, 527-28):

There is a close relationship—if not total identity— Although systems of classification and categoriza- between theories of classification and theories of concept. tion are both mechanisms for establishing order The class of “waterfowl,” for example, includes the sub- through the grouping of related phenomena, fun- classes ducks, geese, and swans, in exactly the same way damental differences between them influence how that the concept “waterfowl” includes the subordinate that order is effected—differences that do make a concepts of ducks, geese, and swans. The different theo- difference in the information contexts established by ries of how we classify birds correspond to the theories each of these systems. While traditional classification of how we conceptualize birds (see Andersen et al. 2006, is rigorous in that it mandates that an entity either is 19-33). Henry Bliss (1929, 120; italics in original) also em- or is not a member of a particular class, the process phasized this: “It is evident that a discussion of classes in- of categorization is flexible and creative and draws volves the correlation of classes to concepts, or class- nonbinding associations between entities—associa- concepts. The class-concept is the mental correlate of the tions that are based not on a set of predetermined class, the mental basis both of the general idea of the class principles but on the simple recognition of similari- and of its name, or names.” ties that exist across a set of entities. Classification Spiteri (2008) found that an examination of traditional divides a universe of entities into an arbitrary system similarity-based concept theories suggests that they do not of mutually exclusive and nonoverlapping classes provide an adequate account of conceptual coherence. that are arranged within the conceptual context es- Library and information science needs to explore knowl- tablished by a set of established principles. The fact edge-based approaches to concept formation, which sug- that neither the context nor the composition of gest that one’s knowledge of a concept includes not just a these classes varies is the basis for the stability of representation of its features, but also an explicit repre- reference provided by a system of classification. In sentation of the causal mechanisms that people believe contrast, categorization divides the world of experi- link those features to form a coherent whole. Spiteri ence into groups or categories whose members bear (2008) found that rather than representing a universal some immediate similarity within a given context. 100 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

That this context may vary—and with it the compo- These two sources pose a question about the broad defini- sition of the category—is the basis for both the tion of classification mentioned in section 2 of the pre- flexibility and the power of cognitive categorization. sent article. Nevertheless, this broad definition is widely used in the literature and it will introduce problems to re- Jacob’s distinction is based on the narrow meaning of strict the term “classification” to the narrow definition. classification presented in section 2. Her distinction is Therefore, the choice made here is to consider classifica- supported by Schmidt and Wagner (2004), who intro- tion as synonymous with categorization but to maintain duced some distinctions between classification and other the distinction between classification in a wide and a nar- forms of discriminatory practice (45-46): row sense.

The point we want to make is that we have to be 3.3 Ordering quite specific in distinguishing different types of discriminative practice: seeing something, seeing WordNet 3.1 provides two senses of the noun “ordering”: something for what it is as opposed to something else (reflecting on what one is seeing), physically – ordering, order, ordination (logical or comprehensible separating things in some regular way, saying that x arrangement of separate elements): “we shall consider is C (“categorizing” x as C), and classifying x as C these questions in the inverse order of their presenta- according to an inscribed, publicly available classifi- tion” cation system. These are radically different practices, – order, ordering (the act of putting things in a sequential involving radically different forms of convention, arrangement): “there were mistakes in the ordering of principles of abstraction, etc. items on the list.”

About categorization, the same authors wrote (391-2): Some authors do not consider historicist classifications (like cladistics systems) as following the concept of classifi- Categorization, by contrast [to seeing and recogniz- cations (Mayr and Bock 2002, 172): ing], is a linguistic operation of ascribing a category or concept to a particular phenomenon by the For several centuries all ordering systems were means of signs. Merely talking about phenomena, thought to be classifications and the two terms were however, is not necessarily categorizing them, al- treated virtually as synonyms. Eventually, however, it though talking involves the application of concepts. was realized that classification means making classes To categorize is to make a conceptual proposition and that ordering systems that are not based on (“red is a color”). classes, such as sequential listing or cladifications In categorizing what you see as trees and birds you (Mayr 1995), are not classifications. Hence, ordering emphasize certain aspects of the world while ab- systems denotes the general concept that includes stracting from others, for instance that the trees and classification as one of its subdivisions. birds may all have green colors or that clouds and leaves may all be moved by the wind. An act of Instead, Mayr and Bock suggest that cladistics systems categorization cuts the world into pieces in that it should be considered as ordering systems in a broader emphasizes certain features at the expense of others category. However, this terminology is not generally used, (“x belongs to category C”). and it is deviant from the suggestions made in the pre- (In themselves acts of separating objects are not sent article. acts of categorization, as they are not necessarily Ordering depends on conceptual classification but it is linguistic operations. Peeling onions or removing broader than systematic classification. Books can be or- dirt from one’s body by means of soap and water dered by, for example, by size, language, or publication are acts of separation but not acts of categorization, date, or alphabetically by author or title. although they may be subjected to acts of categori- For further information about order and ordering sys- zation, for instance when one is instructing children tems, see Meinhardt et al. (1984), Mayr (1995), Mayr and in how to do it. Similarly, when sorting the garbage Bock (2002) and Schmidt and Wagner (2004). (putting paper in this container, potato peels in that container) one may, or may not, be following in- 3.4 Taxonomy structions involving categorizations). The term “taxonomy” was first used in 1813 by the French naturalist Augustin-Pyramus de Candolle (Candolle 1813). Knowl. Org. 44(2017)No.2 101 Reviews of Concepts in Knowledge Organization

It became widely used in relation to biological classification context in which one concept or phenomenon but has since spread to other domains. Some authors con- might be studied within the document. sider it synonymous with classification, whereas others make distinctions between the two terms. In the following This quotation from Slavic is, however, contradicted by quote, the two terms are considered synonymous (Grove other uses of the terminology. Bibliographical classifica- 2010, 5139; references omitted): tions may be phenomenon classifications (for example, the system by James Duff Brown (1862-1914); cf. Beghtol Taxonomy has acquired a wide range of meanings 2004) and disciplinary-based library classification systems no longer restricted to the classical understanding of like the Dewey Decimal Classification (DDC) are also some- biology. Taxonomy is now applying its early sense of times termed taxonomies (see Waltinger et al. 2011). organizing things in accord with particular principles Carl E. Landweh et al. found that a taxonomy is based (“taxis”: arrangement; “nomos”: law) to a broader on a theory (1994, 214): range of domains after several centuries of limitation to biology and other natural sciences. In the 1990s, A taxonomy is not simply a neutral structure for taxonomy was redefined as any semantically signifi- categorizing specimens. It implicitly embodies a the- cant, systematic organization of content or as the ory of the universe from which those specimens are process of developing such organization. This defi- drawn. It defines what data are to be recorded and nition sometimes includes any collection whose indi- how like and unlike specimens are to be distin- vidual elements have been assigned to various nodes guished. In creating a taxonomy of computer pro- of a classification system. Thus, taxonomy is some- gram security flaws, we are in this way creating a the- times considered the process of matching collection ory of such flaws, and if we seek answers to particu- items with predefined labels, and sometimes it is the lar questions from a collection of flaw instances, we creation and arrangement, as well as the resulting must organize the taxonomy accordingly. product, of the classification system itself. However, classifications, too, are based on theories (and Hedden (2016) also uses the term “taxonomy” in a very an atheoretical classification or taxonomy may be consid- broad sense, not just about classifications (hierarchical or ered an oxymoron; see Hjørland 2016b). Therefore, the non-hierarchical) but as a synonym for any kind of knowl- theoretical basis cannot be used as a criterion for distin- edge organization system (KOS). However, one may ask, if guishing classification and taxonomy. the term “taxonomy” is not used with a specific meaning Marradi suggested the following distinctions (1990, in relation to classification and KOS, why then use it at all? 146): According to Aida Slavic (2000), the difference between classifications and taxonomies is based on the distinction A taxonomy obtains when several fundamenta divi- between aspect classification (or “disciplinary classifica- sionis [criteria of division] are considered in succes- tion”) on one side and entity classification (or “phenome- sion, rather than simultaneously, by an intensional cl. non classification”) on the other: [classification]. The order in which fundamenta are considered is highly relevant: the taxonomy obtained Knowledge classification can be, and often is, by using property X to classify a genus and then TAXONOMIC (sometimes called “entity classifica- property Y to classify its species is by no means the tion”) like the classification of zoology, classification same as that obtained by considering property Y first of plants, or classification of chemical elements and property X afterwards. (which means that they are going to list one concept in one place only in the classification structure). Marradi suggested the following differences between Bibliographic classifications i.e. those one has to use classifications, typologies, and taxonomies as products to describe real documents ARE NOT and CAN (129): NOT be taxonomic. They are by all means ASPECT or disciplinary classifications. This means that they 1. Classification schemes will list one concept in all disciplines and fields where When only one fundamentum divisionis is consid- that concept might be studied: e.g. “water” will have ered, a classification scheme is produced—usually to appear under chemistry, physics, in geology, medi- by an intensional classification. The extensions of cine, sport etc. each class must be mutually exclusive, and jointly This is of critical importance for information re- exhaustive. Classes need not be at the same level of trieval as aspect classification helps to establish the generality, and may be ordered. 102 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

2. Typologies A hypothetical fourfold typology When several fundamenta are jointly considered, a Motivated Unmotivated typology is produced. This may be done through ei- Intelligent Success Underachiever ther intensional or extensional classification. 1 2

Unintelligent Overachiever Failure 3. Taxonomies 3 4 When several fundamenta are considered in succes- sion through a series of intensional classifications, a Table 2. A hypothetical fourfold typology (after Bailey 1994, 5). taxonomy is produced. Specific concepts/terms (such as taxon, rank, clade) are needed to deal with The term typology is used in many fields. For example, Carl taxonomies. G. Jung’s psychological types are famous (Jung 1971). In library and information science (LIS) typology is used, for It is not difficult to find examples of the use of the terms example, about document typologies. Web of Science, for “classification,” “typology,” and “taxonomy” in disagree- example, distinguish among article, book review, letter, re- ment with Marradi’s definitions. Here, it will not be dis- view, proceeding paper and other types of documents. cussed whether or not it is a good idea to use his definitions prescriptively. 4. Theories of classification2

3.5 Typology. In this section, the following theories are presented.

“Typology” is derived from the two Greek words (typo) 4.1. The “classical view” of classes as defined by sets of meaning “type” and (logos) meaning “word.” The word elements with necessary and sufficient attributes versus typology literally means the study of types (subdivisions of the views proposed by, derived from, or related to particular kinds of things). We saw above (3.4) that accord- Ludwig Wittgenstein’s late philosophy. As opposed to ing to Marradi (1990) typologies are kinds of classifications necessary and sufficient attributes, these views consider in which more than one fundamental criterion of division that classes and concepts are graded structures. This is simultaneously taken into account. Another definition section briefly presents the prototype theory suggested was provided by Kenneth D. Bailey (1994, 4; italics in by Eleanor Rosch, as well as theories developed by original): Thomas Kuhn and Michael Billig. 4.2. The way of forming classes (e.g., by logical division, Typology is another term for a classification. Two by measuring similarity among elements, by collecting characteristics distinguish typologies from generic elements with a common ancestry, or by collocating classifications. A typology is generally multidimensional tools to support human activities) (the epistemology of and conceptual. Typologies generally are characterized classification).3 by labels or names in their cells. 4.3. The view that there is one correct or best classification versus the view that there are different classifica- Bailey exemplifies: tions for different purposes (the metaphysics of classification). As a hypothetical example, let us use two dimen- sions to construct a classification. These dimen- 4.1 The “classical view” versus “prototype theory”4 sions are intelligence (dichotomized as intelligent/ unintelligent) and motivation (dichotomized as mo- Aristotle developed a theory of classification in which all tivated/unmotivated). Combining these two dimen- elements in a given class share at least one characteristic sions creates a fourfold typology; as shown in Table with all other members. Classes should be designed so 2. These four categories can be defined as cells in membership of a class is given by a set of necessary and the table. In this case, they are types, or type concepts. sufficient characteristics. For example, according to Aris- A motivated and intelligent person can be labeled totle’s per genus et differentiam definition, man is a rational as successful; an intelligent but unmotivated person animal. This definition first considers a class or concept is likely to be an underachiever; while a motivated consisting of all animals (including humans). It then claims but unintelligent person is an overachiever; and one the essential difference between humans (men) and all who lacks both intelligence and motivation is likely other animals is that humans are rational. In this way, the doomed to failure. class of animals is divided into two non-overlapping classes: rational animals (humans) and non-rational animals Knowl. Org. 44(2017)No.2 103 Reviews of Concepts in Knowledge Organization

(all other animals). In order for some organism to belong This classical theory was not the result of empirical to the class of humans, it is necessary that it is rational: all study. It was not even the subject of major debate. It elements in the class must have this characteristic. It is also was a philosophical position arrived at on the basis a sufficient condition: if an organism has the attributes of all of a priori speculation. Over the centuries it simply other animals plus the attribute of being rational, it must became part of the background assumptions taken be human (it is quite a different task to find out if a given for granted in most scholarly disciplines. In fact, until organism is rational or not). This has also been called very recently, the classical theory of categories was monothetic classification5 or the feature theory of classifi- not even thought of as a theory. It was taught in most cation (for further information about the classical view see, disciplines not as an empirical hypothesis but as an for example, Parry and Hacker 1991; Moss 1964; Frické unquestionable, definitional truth. 7 2016). In the middle of the twentieth century, a new theory Geoffrey C. Bowker provided another description (1998, of classification was proposed, which considered itself an 256): alternative to the feature theory that had existed for mil- lennia. Wittgenstein (1953) claimed that not all concepts An Aristotelian classification works according to a consist of elements which have a set of necessary or suf- set of binary characteristics, which the object being ficient characteristics. He used the metaphor of family re- classified either presents or does not present. At semblance for a series of overlapping similarities, where each level of classification, enough binary features no one feature is common to all the elements in the con- are adduced to place any member of a given popula- cept. In a family, some members may be alike in one re- tion into one, and only one class .… Aristotelian spect (e.g., the shape of the nose) while other members models … have traditionally informed formal classi- may be alike in other respects. Wittgenstein’s famous ex- fication theory in a broad range of sciences, includ- ample is games, where he claimed that no characteristic ing biological systematics, geology, and physics. common to all kind of games exists. Experimental psy- Rosch’s (1978) prototype theory argues that, in daily chologist Eleanor Rosch (1978), inspired by Ludwig Witt- life, our classifications tend to be much fuzzier than genstein’s later philosophy, first defined prototype theory.6 we might at first think. We do not deal with a set of Given the Roschian theory, some elements are better rep- binary characteristics when we decide that this thing resentatives than others. For example, if the class or con- we are sitting on is a chair. Indeed, it is possible to cept is “bird,” in the classical theory this concept may be name a population of objects that people would in defined by attributes such as feathers, beak, and the ability general agree to call chairs that have no two binary to fly, and every bird is as representative as any other. In features in common. the prototype theory, on the other hand, a blackbird is considered a good example (at the least by Westerners), There are still many people subscribing to the classical while a penguin is considered a bad example. Instead of view, and Wittgenstein’s view—and thereby, indirectly, being defined by necessary or sufficient characteristics, prototype theory—is criticized by Needham (1975), Sut- classes are determined by the overall likeness to a proto- cliffe (1993) and Margolis (1994). See Fox (2011) for a re- type—hence the name of the theory. cent discussion of prototype theory in knowledge or- A now well-established division between two theories ganization. of concepts and classification is therefore classical or Ar- Thomas Kuhn8 is well known for his book The Structure of istotelian classification on the one side and prototype the- Scientific Revolutions (1962) in which he made the concepts ory (or polythetic classification) on the other. George La- of “scientific paradigm” and “paradigm shift” wide- koff wrote about these two theories (1987, 6; emphasis in spread—including in common language. It is less well original): known that his later research focused on concepts, classification, and scientific taxonomies and inspired a new theory From the time of Aristotle to the later work of of concepts called “theory theory.” The best introduction Wittgenstein, categories [and classification] were to this work is probably that by Andersen et al. (2006, 20- thought to be well understood and unproblematic. 21), who wrote: They were assumed to be abstract containers, with things either inside or outside the category. Things According to this theory, the basic conceptual struc- were assumed to be in the same category if and ture of science is a classification system that divides only if they had certain properties in common. And objects into groups according to similarity relations. the properties they had in common were taken as The grouping is not determined by identifying nec- defining the category. essary and sufficient conditions, but by learning to 104 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

identify similarities and dissimilarities between the species which have becoming similar by adapting to objects. It was one of Kuhn’s central claims that one the same ecological niche. In the future, we may there- learns such concepts by being guided through a se- fore expect that not just experts, but also amateur or- ries of encounters with objects that highlight the re- nithologists, schoolchildren, and the rest of us learn to lations of similarity and dissimilarity currently ac- classify birds in a new way. cepted by a particular community of concept users …. Michael Billig, a social psychologist inspired by rhetoric, Kuhn’s standard example of a learning process of proposed another new theory that is probably in harmony this sort is a child learning the concepts “duck,” with Kuhn’s view in important respects. Billig considers “goose,” and “swan” (Kuhn 1974). In this example, that thinking is like a quiet internal argument. Therefore, an adult familiar with the classification of waterfowl psychological and rhetorical theories are closely linked, guides a child (“Johnny”) through a series of osten- and the psychology of classification/categorization can sive acts until he learns to distinguish ducks, geese, learn much from rhetoric. Billig (1996) describes modern and swans. Johnny is shown various instances of all cognitive psychology’s tendency to consider categorization three concepts, being told for each instance whether a fundamental cognitive process in both animals and hu- it is a duck, a goose, or a swan. mans. Modern cognitive psychology tends to view “the individual as an active processor of information” in which The most important aspects of Kuhn’s theory are: “the effect of a stimulus depends on how it is categorized and interpreted by the perceiver” (quotations from Eiser – People learn concepts (or classifications) according to 1980, 8). This basic psychological process is often attrib- how these concepts are understood in a given society uted a biological status by cognitive psychologists (this by being confronted with exemplars and similarities as criticism is also raised against Eleanor Rosch). Billig finds well as dissimilarities compared with other concepts that the implication of this view is that humans are tied to (e.g., by parents and teachers). prejudiced and bureaucratic modes of thinking. Billig does – Two people can correctly identify the same con- not consider it wrong that categorization is an important cepts/classes even if they use different characteristics process, but from his studies in rhetoric he argues that to make the correct identification. there must be two fundamental processes: categorization – Dissimilarity plays as important a role as similarity in and particularization, the latter being a reverse process in classification. Similarity alone is not enough (see An- which something is not just considered an element of a dersen et al. 2006, 24ff). class or a category, but is considered something special. – A given concept/classification is based on a paradigm. Billig further demonstrates with many examples how hu- For example: mans are able not just to categorize and particularize but – Paradigm one: Ptolemaic astronomers might learn the also to discuss and consider the way things are catego- concepts “star” and “planet” by having the Sun, the rized. Arguments about categories and particulars are im- Moon, and Mars pointed out as instances of the con- portant elements in human communication and thinking, cept “planet” and some fixed stars as instances of the and they are often related to wide-ranging theoretical and concept “star.” ideological issues and conflicts. By including particulariza- – Paradigm two: Copernicans might learn the concepts tion as a basic psychological process, Billig is able to make “star,” “planet,” and “satellites” by having Mars and room for people, who are not just prejudiced and bureau- Jupiter pointed out as instances of the concept cratic but also open-minded and flexible. In relation to re- “planet,” the Moon as an instance of the concept “sat- search in knowledge organization, Billig’s research raises ellite,” and the Sun and some fixed stars as instances of serious problems for the cognitive view that tries to base the concept “star.” Thus, the concepts “star,” “planet,” classifications on the study of the human mind. and “satellite” got a new meaning and astronomy got a The basic lessons from these new theories of classifi- new classification of celestial bodies. cation may be summarized this way:

The difference before Copernicus, and later, say, Newton, a) Humans do not classify in a given way according to is striking: after a paradigm shift we learn to distinguish inborn or “given” characteristics, but according to the concepts in new ways (see Andersen et al. 1996). human activities and goals that we have (which may require different classifications). – A contemporary example: Ornithologists have recently b) Instead of the classical model of sets of mutually ex- discovered that the blackbird, which so far has been clusive and jointly exhaustive criteria, we may need al- considered one species, should be considered different ternative models. Andersen et al. (2006) found that the Knowl. Org. 44(2017)No.2 105 Reviews of Concepts in Knowledge Organization

so-called “dynamic frames” represent the best way to tion words, such as adverbial particles and preposi- represent classes. It has not been examined, however, tions. whether this form of knowledge representation is applicable or fruitful in relation to the construction of – A third problem with the picture theory is that it knowledge organization systems (KOS) in LIS. represents knowledge of the world as the conjunction of knowledge of independent microworlds. To regard 4.2 The methodology of forming classes, the totality of knowledge as a simple aggregation is the epistemology of classification simplistic ….

The fundamental elements of any classification are its theo- Concerning (γ), Svenonius’ last theory was the contextual retical commitments, basic units and the criteria for ordering or instrumental theory of meaning. The basic tenet of the these basic units into a classification—Hull 1998 instrumental theory of meaning is that we know what a word means when we know how to use it. Svenonius 4.2a Elaine Svenonius (2004) proposed that three epis- found that this way of thinking led to adoption of the temological theories are important for knowledge organi- methods of numerical taxonomy. However, numerical tax- zation (or, as she preferred, knowledge representation): onomy may be considered a form of empiricism rather than of the contextual or instrumental theory of meaning α Operationalism as developed by pragmatic philosophers. If this under- β The picture theory of meaning standing is true, we may conclude that Svenonius has not γ The contextual or instrumental theory of mean- really suggested an alternative to empiricism and logical ing positivism. Such alternatives are presented below (4.2c). First, however, another important view will be presented. These views may deserve their own entry in this encyclopedia. They are outlined below. 4.2b Alberto Marradi distinguished the following senses of classification as an operation (intellectual and otherwise): Concerning (α), Svenonius raised the criticism that all op- erational deﬁnitions lack validity and that operationalism α Intensional classification (or subdivision or represents a form of logical positivism. downwards classification)9 The subdivision of the extension of a concept Concerning (β), what Svenonius termed “the referential (genus) into several extensions corresponding to or picture theory of meaning,” she found that this, too, as many concepts of lower generality (species). “derives from an empiricist view of knowledge” (2004, The former and all the latter concepts have the 574). The author summarized the basic problems with same intension except for one aspect (fundamen- this theory (578, note omitted): tum divisionis): on that account each species concept is a different partial articulation of the ge- – First, the picture theory assumes a universal form of nus concept (Marradi 1990, 129; emphasis in language in which the meaning of propositions pictur- original). ing the world are prescribed, relatively fixed, and gen- β Extensional classification (or numerical taxonomy erally understood. The objection here is that pictures or upwards classification)10 can be differently interpreted. A cup is half full or half The grouping of the objects/events of a set into empty. A picture of a duck from another viewpoint several subsets according to the perceived simi- could be a picture of a rabbit; a picture of a block larities of their states on one or more properties could be interpreted as a triangular prism. (Marradi 1990, 129). Botanist Michel Adanson stated that “all parts – Secondly, the picture theory implies fixity of reference. and qualities, or properties and faculties of But the meanings of words are not necessarily fixed in plants... barring not even one” ought to be con- the sense of referring to sets of homogeneous objects sidered before attempting a classification (1763, in the real world or clearly delineated mental concepts. clvi). Along with this idea, Adanson operated ex- Many words have fluid boundaries. (A chair with three tensional [classification] and produced taxono- legs is still a chair.) Fluidity is necessary if words are to mies based on the rate of equal states on the to- function in a variety of different contexts. The picture tal of properties considered between any two theory falls down particularly in the case of abstract plants (1763, vol. I) (Marradi 1990, 136). words whose referents are mental constructs and func- 106 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

The predominant criterion is to maximize ho- cation] had to wait for the development of another mogeneity within classes and heterogeneity be- intellectual tool, viz. the idea of orderly recording tween classes (Marradi 1990, 135). the states of a vector of objects on a vector of Other labels have been proposed for the opera- properties—in other words, for the intellectual tion, over and above the old ones—“classifica- forefather of what is presently known as the data tion” and “taxonomy.” Among them “numerical matrix. taxonomy” (Sokal and Sneath 1963), “class formation” (Capecchi and Möller 1968), “cluster What are the major benefits and drawbacks of logical di- analysis,” etc. (Marradi 1990, 136). vision as a method of classifying? Frické (2016, 547) γ Classing stated: “Logical division produces classifications with The assignment of objects/events to classes de- admirable qualities. Everything has a place in a leaf, its fined by the first operation [subdivision] (or of own unique place, and the classification schedule embod- new objects/events to groups created by the ies the maximum amount of general information about second operation with other objects/events) the items being classified.” Its weaknesses have been (Marradi 1990, 129). known for a long time: “Aristotle had argued that logical division was an inappropriate tool for the classification of Marradi’s two first options are further discussed in the organized beings” (Stevens 1998). One limit of this next section. His third option is not related to a new fun- method is that it seems better suited to some kinds of damental method and is not further discussed in this arti- object (e.g., formal objects, such as mathematical objects) cle, but this issue is partly dealt with in Hjørland (2017). compared to other kinds of object (e.g., “organized beings”), but this is an open issue today. A modern criticism 4.2c Birger Hjørland has suggested that there are four ba- is its relationship with essentialism. If the basis of divi- sic theories and approaches to classification11: α rational- sion (fundamentum divisionis: classification principle) is ism; β empiricism; γ historicism, and δ pragmatism/critical not to be arbitrary, it has to be deduced from what are theory. All four will be presented and discussed below. The considered essential criteria, but this idea is heavily criti- first two (rationalism and empiricism) are related to theo- cized today (Wilkins 2013; Frické 2016). ries already presented. 4.2cβ Empiricism. The second of Marradi’s operations, 4.2cα Rationalism. Hjørland considers than the first of extensional classification, has, according to Marradi, also Marradi’s operations, intensional classification or subdivi- been termed “numerical taxonomy” and “cluster analy- sion, corresponds to what he in different writings has re- sis,” among others. It corresponds to what Farradane lated to rationalism (2011, 74): (1950) termed inductive classification and Parrochia [2016] “phenomenal classifications,” and has by Hjørland Rationalist theories of indexing (such as Rangana- been related to empiricism (2011, 74): than’s theory) suggest that subjects are constructed logically from a fundamental set of categories. The Empiricist theories of indexing are based on the basic method of subject analysis is then “analytic- idea that similar (informational) objects share a synthetic,” to isolate a set of basic categories large number of properties. Objects may be classi- (=analysis) and then to construct the subject of any fied according to those properties, but this should given document by combining those categories ac- be based on neutral criteria, not on the selection of cording to some rules (=synthesis). The application properties from theoretical points of view because of rules such as logical division is by principle part this introduces a kind of subjective criteria, which of the rationalist view. is not approved by empiricism. Numerical statistical procedures are based on empiricist philosophy. According to Marradi, logical division dominated for centuries until challenged (or supplemented) by extensional This means that overall likeliness, sometimes termed (empiricist) classification (1990, 135-36). “phenetics” (e.g., statistical measures of similarity based on a great number of attributes), is the basis of this method. In our opinion, this belated development [of em- This seems at first to be properly scientific but, on further piricist classification] depends on the fact that, in examination, it turns out to be based on a number of order to be somehow formalized from a spontane- problematic assumptions. The first question is about the ous activity into a respectable intellectual operation number of properties needed (Marradi 1990, 137). within a scientific discipline, extensional cl. [classifi- Knowl. Org. 44(2017)No.2 107 Reviews of Concepts in Knowledge Organization

A controversial question is how many properties number of skeletal similarities. Different methods and are to be considered. Parsimony and elegance paradigms in biological taxonomy thus arrive at different would advise to keep that number down; on the results: Methods based on historical development consider other hand, it may be remarked that “increasing the birds and reptiles as related species (birds may be consid- number of variables increases the probability of ered a kind of reptile), while numerical taxonomy, based on correct classification” (May 1982, 43). Since the a quantitative study of many structural similarities, consid- concept of “correct [classification]” is rather ques- ers birds and reptiles to be less related (birds are not rep- tionable …, a better argument might be that, with a tiles). The reason is that many animals develop similar formalized approach, there is no way to consider structures because they adapt to the same ecological the information on the objects’/events’ properties niches, but may have very different phylogenetic back- unless those properties are included in the matrix. grounds and are by biologists considered different species. Therefore, leaving properties out of the matrix en- (Considering the reverse, in some spiders the male and the tails a loss of information of unknown amount. female are very different, and if numerical taxonomy is used, the male and the female might be classified as differ- If we assume that any object has an unlimited number of ent species, which obviously is problematic.) Therefore, properties, then it is impossible to consider them all, even modern biological taxonomy is dominated by quite a dif- in theory. By implication, any empirical classification is ferent philosophy and method: cladism. We shall return to biased in ways that cannot be controlled. this below. Consider, however, that Marradi saw these two Another problem is that the descriptions of objects, methods (logical division and phenetics) as exhaustive, but on which this method is based, cannot be atheoretical or that Hjørland (2009) presented two additional methods of objective (cf. Hjørland 2016a, 2016b)—or, as formulated classification related respectively to historicism and prag- by Gitelman (2013), “Raw data is an oxymoron.” In other matism/critical theory. The biological examples should words, the data used are always theoretically biased in provide sufficient argument for the first of these, and oth- ways we often cannot recognize or control. ers have argued in a similar way.12 The third problem, closely related to the second, concerns the concept of similarity. Classification has often 4.2cγ Historicist approaches to classification been defined as bringing like things together (and thus separating unlike things). “Likeness” is a concept that To say that two elements belong to the same class (or may also be expressed by other terms such as “similarity,” “clade”) if they share a common ancestor is clearly dif- “sameness,” “resemblance,” or “equivalence.” The prob- ferent from defining membership of a class by similarity lem is that things cannot be similar in an objective way. (sets of characteristics as arranged by logical division or Any object is similar to another object in some ways and numerical taxonomy). Today, this is the dominant ap- dissimilar in other ways. For any three objects, two differ- proach in biological systematics (termed “cladistics,” ent classifications can be constructed which fulfils the “phylogenetic classification,” “historical classification,” demand of bringing the like objects together. Consider “genealogical classification,” or “genetic classification”). Figure 1 below, wherein the items may be classified ac- This approach is based on the historical or evolutionary cording to color or shape. None of those properties is development of the classified objects (Hennig 1966; objectively more important than the other. For some Hjørland 2003, 107; Gnoli 2006). It is not only used in purposes, the two squares are most alike and should be biology, but also, for example, for classification of lan- classified together. For other purposes, the two black fig- guages and musical instruments—and should be consid- ures (a square and a triangle) are most alike and should be ered one of four general approaches to classification. classified together (see also Popper 1959, 441). Hjørland (2013, 2016a) suggested genealogical classification may also be used in bibliometrics and information ■ □ ▲ retrieval as an alternative to classifying documents ac- Figure 1. Classification criteria. Which two are similar? cording to similarity. To define membership of classes, clades, or species by common ancestor is different from, By implication, empirical criteria for classification are not but related to, an evolutionary ordering of classes (see enough and we need some guidance on how to determine Dousa 2009 for early discussion of evolutionary order in which criteria should be used when determining similarity. library classification). If we take a biological example, scientists have long recog- Although cladism seems to dominate biological taxon- nized that modern-day birds and reptiles share a common omy today, it has also been met with skepticism (Dupré ancestor. Both groups lay shelled eggs and have scales (in 2006, 31): birds, confined to the legs), nucleated red blood cells, and a 108 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

It is at the same time becoming clearer that there is and learn to set it aside when considering foreign or very likely no such ideal classification. There is no historical colour descriptions. The aim is to dispose reason why a classification that reflects the origins of of any preconceptions about how colour “should” the things classified should coincide exactly with one be classified and described, so as to gain insights aimed at the ecological relations of those things, and into the workings of other languages and cultures, it is increasingly perceived that these can and do di- and into the nature of colour itself. verge (Dupré 2002, chapters 3-4). This possibility becomes even clearer in view of the difficulties that Formerly, Hjørland wrote (2011, 74): are emerging in the project of evolutionary-based classification. Speciation was once seen as an all or Hermeneutical theories of indexing suggest that nothing affair leading to complete isolation of one the subject of a given document is relative to a group from another. It is now clear that for micro- given discourse or domain and is why the indexing organisms, in particular, there is very little such isola- should reflect the need of a particular discourse or tion, and genetic material moves in many ways from domain. According to hermeneutics, a document is one kind of organism to another. In fact it has be- always written and interpreted from a particular ho- come common to conceive of the genome of an rizon [note omitted]. The same is the case with sys- ecosystem (the soil of an area, or a body of water) tems of knowledge organization and with all users rather than the privatized genome of an individual searching such systems. Any question put to such a organism (e.g., Venter et al. 2004). The classical pic- system is put from a particular horizon. All those ture of speciation applies quite well to some of the horizons may be more or less in consensus or in most complex multi-celled organisms, such as conflict. To index a document is to try to contrib- mammals and birds, though much less well to plants. ute to the retrieval of “relevant” documents by An important movement in biology is to transcend knowing about those different horizons. the anthropocentrism that takes the peculiarities of our own corner of the living world as the model for Historicism is therefore, as we have seen, an approach all. that may be applied to both the object and the subject in classification. If both the object and the subject are con- Dupré’s reservations are not about the validity of the sidered, we may speak of a united historicist theory. B. M. cladistics approach as such, but about the idea of one ideal Kedrow presents such a united historicist view of classi- classification. It seems clear that genetic classification is a fication (1975, v. 1:4-5; translated from German by BH.): distinct approach, with some major benefits—to know about the origin of things is to know things in a deeper Historicism as a key to any natural classification. way than just to know about sets of attributes (which, in Of crucial importance for the analysis of the prob- the historical perspective, often looks superficial). lem at hand is the historical approach to its consid- There is also a subjective side of classification, and this eration and solution, in other words, the principle of subjectivity has developed historically. The classifying sub- historicism. This refers both to the development ject is influenced by his or her culture, paradigms, and tra- history of the objects studied by the sciences as well dition. This may be termed, for example, hermeneutics, as to the evolution of scientific knowledge itself. historicism, or social epistemology. This can be fruitfully il- In fact, any artificial classification of things or lustrated in the classification of colors (Biggam 2015, 1): knowledge of things is mainly characterized by the lack of historicism. In that case, the relationships When the colour vocabularies of various languages between things and between things and the knowl- are considered and compared, the researcher finds edge of them will not reveal the classification as a that there are many different ways in which humans result necessarily incurred in the course of devel- categorize and “label” colours, resulting in an amaz- opment but rather as random and superficial rela- ing array of misunderstandings. Monoglot indi- tions that sometimes are also fixed by man himself viduals invariably believe that their own colour sys- in order to understand the given material. tem is clear and obvious, and they are often mysti- In contrast, the truly natural classification reflects fied when confronted with an alternative system. So the real connections between objects as they have the first step which the reader has to take when en- evolved in their development or in the development tering the world of colour semantics is probably of knowledge of them. This was, for example, the the most difficult of all; s/he must restrict his or case with the classification of the chemical elements. her own colour system to normal, everyday speech, This truly natural system could only be made after Knowl. Org. 44(2017)No.2 109 Reviews of Concepts in Knowledge Organization

the relationships between the elements had been es- classifiers themselves and the characteristics that they tablished as real interactions, which had found static choose can affect classification”15 and they observed “that relations and given tabular form. All previously es- objectivity is neither possible nor desirable in classifica- tablished systems of elements proved to constantly tion. Despite the arbitrariness, some classifications can be be one-sided, in essence artificially, in a sense, arbi- more reasonable or more useful than others” (573). Dif- trary, because none of them were based on the prin- ferent interests and kinds of subjectivity may not be ex- ciple of historicism. plicit (or they may be in conflict with their stated goals) and it is therefore an important task to uncover the hid- A more recent view was expressed by Fulvio Mazzocchi den assumptions in classification principles, classification (2017, 373): criteria, and in all kinds of knowledge organization systems (KOSs) and information retrieval (IR). As formerly Hermeneutics and postpositivist epistemology em- stated (Hjørland 2011, 74): phasized, respectively, the historicity of understanding and the incommensurability13 of alternative sci- Pragmatic and critical theories of indexing are in entific paradigms. Postmodernist theories argued for agreement with the historicist point of view that the breakdown of “grand narratives,” indicating the subjects are relative to specific discourses but em- need to embrace pluralistic views. What is basically phasize that subject analysis should support given common to many of these approaches is the refusal goals and values and should consider the conse- of the belief that an absolute vantage point can be quences of indexing. These theories emphasize that reached. There is no ultimate criterion for univocally indexing cannot be neutral and that it is a wrong distinguishing accidental from distinctive features: goal to try to index in a neutral way. Indexing is an the fixing of such a distinctiveness always depends act (and computer-based indexing is acting accord- on a given perspective. ing to the programmer’s intentions). Acts serve human goals. Libraries and information services [and The most concrete implication of the united historicist classifications] also serve human goals, and this is view of classification is that theories become important in why their indexing should be done in a way that the explanation for our categories compared to similarity supports these. or other criteria (cf., Murphy and Medin 1985; Hjørland and Nissen Pedersen 2005). Different theories or para- In recent years, there has been a focus on ethical issues in digms imply different classifications; therefore to provide knowledge organization (e.g., Adler and Tennis 2013), as design principles for classifications is to negotiate the dif- well as on the consequences of classification (e.g., Bowker ferent theoretical influences on the domain to be classi- and Star 1999) and feminist approaches to knowledge or- fied. Whereas empiricists and positivists tends to “let the ganization (e.g., Fox and Olson 2012). Together with re- data speak for themselves,” the hermeneutics-oriented research, uncovering hidden assumptions in classification searcher tends to apply a broad orientation which is able and arguing about the paradox of atheoretical classifica- to uncover the theoretical influences that have produced tion (e.g., Hjørland 2016b), the pragmatic/critical ap- the data and their interpretations and classifications in a proach is exemplified. given domain. This issue brings us to the last of the basic One may wonder, however, if pragmatic/critical classi- approaches to classification: pragmatic and critical theo- fication is scientific or able to function as the theoretical ries. basis for classification research and practice. The first impression might be that this is a decline in scientific 4.2cδ Pragmatic and critical approaches progress. For example, to classify animals in relation to to classification human interests as domestic animals, pets, and pests seems primitive compared with biological classification in The pragmatic and critical approach to classification is which no such interests and goals seem to be used. The based on considering the goals, values, interests, policies, answer to this argument is that there are levels of prag- and consequences of classification. There may be many matic classification and that basic science can be inter- different values at play in forming classifications.14 Classi- preted from the perspective of pragmatic philosophy. fications based on this approach are constructed in order Pragmatism may be more or less short term or long term, to support explicit interests. From this perspective, a clas- and the pragmatic value of cladism should be understood sification can never be neutral, but will always tend to from the perspective of long-term interests. Critical the- support certain goals and interests at the expense of other ory claims that, in the end, our scientific theories should interests. Nobes and Stadler (2013) examined “how the be evaluated from the perspective of human practice. 110 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

One of the reasons for considering pragmatism/critical minological distinction between artificial and natural sys- theory as a serious approach is that the other approaches tems, and this was praised as one of his main achievements (rationalism, empiricism, and historicism) have problems by later naturalists and philosophers” (Müller-Wille 2007, that are not less significant. We cannot have a science of 550; cf. Müller-Wille 2013, 311). classification without considering criteria for classification Although Linnaeus considered natural classification the as the fundamental problem. The claim of pragmatism ideal, he recognized that his own system (at least partly) and critical theory is that, in the end, such criteria must represented an artificial classification (Stevens 1998): serve human values and interests (see Pihlström 2009 for scholarly argumentation). It is extremely important to re- Linnaeus realized that natural orders could not be alize, however, that truth is always the goal in science and defined. Even the most “natural,” such as the Um- scholarship. Worst of all is the manipulation of research belliferae, the carrot family, lacked features that were in order to serve some specific interests. Like the histori- unique to and constant within them. Until these were cist approach, the pragmatic/critical approach bases de- found, natural groups were “like a bell without a sign principles for classification on the negotiation of the clapper”; in modern parlance, they were polythetic. different theoretical influences on the domain to be classified, but it provides some additional criteria for theory However, the meaning of “natural classification” has of- analysis and evaluation, such as the social conditions un- ten been considered unclear (Stevens 2016, 494): der which knowledge is being produced. In the middle of the seventeenth century, many, like 4.3 Some metaphysical issues of classification Aristotle 2000 years before, believed in a nature that (is there one correct classification?) could be represented as some version of the scala naturae, a linear sequence of organisms arranged Has the world one unique structure (“taxonomic mo- according to ideas of “highness” and “lowness,” in nism”), or is there more than one structural entity and which man was above all organisms (and often not process (“taxonomic pluralism”)? Are the structures of part of nature), and angels and ultimately god [sic] the world mind-independent (realism), or are they arte- might be above him. There were many other ways facts projected into the world (idealism)? Can our classifi- of representing nature, and as the geologist Francis cations be natural, or are they always artificial? These are Bather observed in 1927, “not a single naturalist had core issues in the metaphysics of classification. As stated a clear idea of what he meant by “natural.” All he by Anjan Chakravartty (2011, 157): knew was that the other fellow’s classification was unnatural” (Bather 1927). In early usage, natural his- The preeminent question of the metaphysics of clas- tory itself for the most part had no historical ele- sification is that of whether the world is itself natu- ment; “history” meant “story” or “description.” rally subdivided into kinds of things. Are kinds out there, so to speak, or are they rather artefacts of John Stuart Mill suggested the following definition (1872, convention, existing only insofar as classificatory 498): practices are brought to bear by creatures such as ourselves? The Linnæan arrangement answers the purpose of making us think together of all those kinds of We shall here present the following issues: plants, which possess the same number of stamens and pistils; but to think of them in that manner is a Artificial versus natural classification of little use, since we seldom have anything to af- b Order versus disorder of things and unity versus dis- firm in common of the plants which have a given unity of knowledge number of stamens and pistils. The ends of scientific classification are best an- 4.3a Artificial versus natural classification swered, when the objects are formed into groups respecting which a greater number of general proposi- Natural classification can be expressed by Plato’s metaphor tions can be made, and those propositions more of “carving nature at its joints” (Plato c.370 BC, Phaedrus important, than could be made respecting any other 265e; see also Campbell et al. 2012). Carl Linnaeus is, how- groups into which the same things could be distrib- ever, often recognized as the first scholar to clearly have uted … differentiated “artificial” and “natural” classifications: “As A classification thus formed is properly scientific or far as I can see, Linnaeus was the first to draw a clear ter- philosophical, and is commonly called a Natural, in Knowl. Org. 44(2017)No.2 111 Reviews of Concepts in Knowledge Organization

contradistinction to a Technical or Artificial, classi- that competing theories provide competing natural classification or arrangement. fications. There are two additional aspects to consider in relation Alphabetical arrangement is also a kind of artificial classi- to natural classification: the concepts of “natural kind” fication because it is not based on relationships between and “naturalism in classification.” objects, but on formal characteristics of the names of ob- A natural kind can be understood as a grouping that re- jects. Artificial classifications often serve important but flects the structure of the natural world rather than the in- limited practical purposes, whereas natural classifications terests and actions of human beings. Chemical elements— have broader application functions. (A yet more limited e.g., gold—are often taken as an example of a natural kind. kind of artificial classification is “ad-hoc classification,” However, the philosophical problems of natural kinds, and which is just made for a specific task, non-generalizable, how it can be decided if something is or is not a natural and not intended to be able to be adapted to other pur- kind, are big ones (obviously, the social constructivist de- poses: cf., Hudon et al. 2005). Marradi found, however, nies that natural kinds exists). Here we shall not go further that use of the term “natural classification” is often con- into this issue but leave it to a hopefully forthcoming inde- nected with a problematic positivist view.16 pendent article. There is a large literature on this concept: William Parry and Edward Hacker explained the con- see, for example, Khalidi (2013). cept in accordance with John Stuart Mill’s understanding: Naturalism in classification may be understood as a general approach to classification theory that establishes a For example, one may divide rocks—or even ani- close connection between knowledge organization and mals—into those weighing less than ten grams, classification in empirical science and scholarship (e.g., bio- those weighing at least ten but less than twenty logical classification, classification of the chemical and grams, and so on; but this is likely to be of little physical elements, classification in arts, linguistics, psychia- use, except perhaps for knowing what it would cost try, etc.). Naturalism is therefore opposed to the idea that to mail them (Parry and Hacker 1991, 133). the field of knowledge organization has a set of a priori classification principles or methods. Naturalism in classifi- And later: cation is based on the corresponding concept of “naturalistic epistemology,” which has been described the follow- [A classification] is fruitful to the extent that it sug- ing way (Rysiew 2016): gests new hypotheses, explanations, and theories concerning its subject matter. For example, the pe- Broadly speaking, however, proponents of NE riodic table—the classification of the elements— [naturalistic epistemology] take the attitude that there proved extremely fruitful, since it suggested the ex- should be a close connection between philosophical istence of hitherto unknown elements and even investigation—here, of such things as knowledge, suggested what physical properties they would justification, rationality, etc.—and empirical (“natu- have. It should be noted that natural classifications, ral”) science. by definition, are more fruitful than artificial ones (Parry and Hacker 1991, 139). Naturalistic classification is therefore the attempt to learn classificatory principles by studying how the most success- Hjørland (2016b) considered the classification of mental ful classifications have been constructed in different do- disorders in the DSM system. The third edition of this mains, as well as the discourses and controversies about system especially claimed to be atheoretical and tended to classification and its philosophy. In the field of knowledge give priority to reliability in diagnosis rather than in the organization, this has been relatively neglected because the validity of classifications. This creates a system with field has tended to provide prescriptive principles on how doubtful functions in the understanding and treatment of to classify knowledge (e.g., facet analytic principles, stan- mental diseases. While it is relatively easy to make a classi- dards for thesaurus construction, or user-based method- fication reliable (e.g., by classing according to weight, as ologies). in Parry and Hacker’s quotation), it is much more difficult, but also much more important, to make a classifica- 4.3b Order versus disorder of things and unity tion useful for predicting the outcome of interventions. versus disunity of knowledge Therefore, the distinction between artificial and natural classification is important when natural is understood as a It has been (and probably still is) characteristic of many classification based on a substantial theory—implying researchers to believe in a fundamental order underlying the apparently confusing empirical picture of the world. 112 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

Related to this view is a belief in the nature of knowledge This paper traces and interrogates the shift from to reflect or converge toward this underlying order. The classification-as-ontology, in which everything is de- first point is about taxonomic monism versus pluralism; fined as it is, to a more contemporary notion of the second is about descriptive monism versus pluralism classification-as-epistemology, in which everything (or epistemic relativism). Often taxonomic monism is as- is interpreted as it could be—or more precisely, the sociated with scientific realism, while taxonomic plural- paper argues for a conceptual move from modern ism is associated with relativism. Chakravartty (2011) ar- monistic ontology to late-modern pluralistic epis- gued, however, that taxonomic monism is in opposition temological foundation for classification theory and to contemporary science and that a form of taxonomic practice. pluralism is consistent with realism. Henry Bliss is a library science representative holding This opens many questions, and the most important the belief in an underlying order of things and in the unity claims by Mai may be considered the critique of the posi- of and consensus of knowledge. He wrote (1933, 37): tivist view that the researcher and knowledge organizer are neutral agents providing an objective mirror of the The more definite the concepts, the relations, and universe of knowledge (see also Pando and de Almeida the principles of science, philosophy, and education 2016). become, the clearer and more stable the order of The same issue was also addressed by philosopher the sciences and studies in relation to learning and Finn Collin, who discussed a social constructivist view of to life; and so the scientific and educational consen- classification, which he formulated in this way (1993, 29: sus becomes more dominant and more permanent. italics in original):

A critique of this view was made by Satija (1992, 40-41), To isolate a certain kind of thing is the same process paraphrasing McGarry (1991, 148): as classifying individual things. And classification is a matter of sorting things into groups, the members Knowledge is a cultural entity and keeps shifting its of which are more similar to each other than to pattern like a kaleidoscope. An emergence of the items outside the group. However, things are only new knowledge modifies the structure of the similar or dissimilar in certain respects …. Classifica- whole. Contrary to H. E. Bliss (1870–1955) there is tions are not objective divisions, inherent in the na- no permanent order in knowledge. “Pattern is new ture of things, but are structures we impose upon every moment,” said T.S. Eliot (1888–1965), with a nature … kinds of things are indeed human crea- poetic vision. tions.

In information science and knowledge organization, this Then Collin wrote (29): “I believe this reasoning is mis- tension between the idea of order and stability and the re- taken. What follows from the premises is a less radical ality of disorder and relativity is clear in the differences be- conclusion.” Later, he summed up his argument (1993, 43; tween traditional bibliographical classifications on the one italics in original): hand, and the bibliometric maps based on, for example, co-citation patterns on the other. Whereas traditional clas- The nominalist argument mistakes a valid anti- sifications tend to provide relatively stable structures, the essentialist point for an anti-realist one. It is true that citation practices of researchers tend to display very unsta- there is not, among the true descriptions of a thing, ble patterns. one which is privileged, in the sense that any classifi- While the goal of knowledge organization is to discover cation of the thing has to be based upon that par- or construe some kind of order, the nature of the order ticular description. There is no uniquely correct clas- matters. Francis Miksa, for example, wrote: “In the end, sification of a thing, one that shows what the thing there is strong indication that Ranganathan’s use of faceted really is, rendering alternative classifications some- structure of subjects may well have represented his need to how misleading or inappropriate. But it is a mistake find more order and regularity, in the realm of subjects, to infer from this that things do not in themselves than actually exist” (Miksa 1998, 73). This quotation may belong to any classifications at all and that things be considered a criticism of Ranganathan in accordance only come to belong to classes when we place them with Hjørland’s (2014) criticism of rationalism. there. Once we relativise similarity and dissimilarity Jens-Erik Mai discussed this from the perspective of to particular aspects of things, similarity and dissimi- post- or late modernist philosophy (2011, 711): larity turn out to be objective, although relational, properties of things, and the predicates that are de- Knowl. Org. 44(2017)No.2 113 Reviews of Concepts in Knowledge Organization

fined by the equivalency classes of things turn out to sition can be listed. Ontologically, there is: (1) a real- be genuine properties of those things. Reality pos- ity that exists independently of us, (2) a single set of sesses all the properties attributed to it in these alter- categories, and (3) a single set of essential properties native descriptions. and therefore a unique way of dividing the world into kinds. Epistemologically, saying that also implic- An issue regarding Collin’s argumentation is “the true de- itly means that: (4) the ultimate order of the world is scriptions of a thing.” Because descriptions are theory- within the reach of human cognitive means, i.e. there dependent, different descriptions are not necessarily is an absolute or neutral vantage point from which to equally true. Also, of course, classifications are made, cho- grasp reality as it “actually” is, and this vantage point sen, or used for a purpose, and therefore our classifications is accessible to us. to a large degree will be human creations (but not there- Such a model has had a strong influence on logic, fore arbitrary or contingent; within ornithology, for exam- philosophy, and science. In Western culture, the tree ple, there seems now to be considerable optimism that a structure based on Aristotle’s logic has been the “final” classification of birds on the overall level seems dominant model of classification. Scientific taxo- within reach; cf. Fjeldså 2013).17 nomic thinking (for instance, the Linnaean classifica- Mazzocchi wrote about the opposite of one right way tion system in biology) and many contemporary se- to “carve nature at its joints,” taxonomic pluralism (2017, mantics theories (for instance, Chomsky’s sentence 373. See also Galison and Stump 1996): diagrams) also embody a similar scheme.

Ontological pluralist views have also been devel- The tree metaphor is increasingly ousted by alternatives oped. For instance, Dupré’s (1993) “promiscuous such as net and the rhizome (see Mazzocchi 2013) and it realism” conceives the world (his argument refers, matters which models guide our research and practice. above all, to the biological realm) as made up of a multidimensional complexity: things are intercon- 5.0 Conclusion nected and interrelated to one another in multiple ways; there is no unique way of carving nature at its The concept of classification and its associated theories is joints or one ultimately right way of classifying or extremely wide-ranging and interdisciplinary. Many kinds hierarchizing. Rather, there are many equally legiti- of knowledge must be combined in order to make overall mate ways of dividing the world into “kinds,” de- progress in this field. Such projects may be guided by dif- pending on the purposes of investigations. ferent basic assumptions. Daniel Parrochia seems to base Returning to Greek philosophy for a moment, we his suggestion for such a program on more formal and find that even Aristotle, with respect to zoological mathematical approaches. In 2016 he wrote: classification, supported a pluralist view (see Parts of Animals I and History of Animals) that partially In spite of these advances, most of classifications resemblances Dupré’s view (Henry 2011). The bio- are still based on the evaluation of resemblances be- logical world contains natural kinds marked by real, tween objects that constitute the empirical data. objective boundaries, but at the same time it is not This one is almost always computed by the means possible to assign animals to a unique set of mutu- of some notion of distance and of some algorithms ally exclusive and non-overlapping kinds. Even in of aggregation of classes. So all these classifications Aristotle’s view, many cross-cutting joints can be remain, for technical and epistemological reasons found in nature. Which joints are chosen to be cut that are detailed below, very unstable ones. A real al- along depends at least partially on the explanatory gebra of classifications, which could explain their context. Since there are diverse explanatory pur- properties and the relations existing between them, poses in zoology, organisms can be grouped into is lacking. Though the aim of a general theory of various (cross-cutting) kinds. classifications is surely a wishful thought, some recent conjecture gives the hope that the existence of In this connection, it is worth mentioning the mental a metaclassification (or classification of all classifica- models or metaphors that govern our view of how tion schemes) is possible. knowledge is organized (Mazzocchi 2017, 372). Alternatively, another program may be suggested (not The tree model has been historically associated with necessarily in conflict with Parrochia’s). This program is the philosophical position of classical realism. A less formal (and therefore more substantial) and views number of basic assumptions underlying such a po- classifications as tied to (domain) theories. By implica- 114 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization tion, the study of classification involves the study of sifications to take over from classic (and too lim- theories in different domains and the ontological claims ited) versions of set theory. of those theories. The justification of a good classification in this perspective is to make a justification of the Parrochia and Neuville (2013) seem to assume that a theoretical premises on which it is based. general theory of classification(s) must be a mathematical theory, but do not discuss if other views may Notes also be important, or what the relative contribution of mathematics is. No doubt mathematics is important, 1. Wesolek (2012, 1) stated: “He [Hjørland] thinks that but it is certainly not all there is to say about classifi- concept classification should not strive to classify on cation. the basis of the properties of objects, but rather on de- 3. An anonymous reviewer wrote: “It is a bit funny to scriptions of objects that are loosely derived from hu- call the methods of forming classes ‘epistemology’ man activity and social negotiation.” In order to classify (which is theory of knowledge). If anything, it would by properties, we have to know these properties and be ‘ontology.’” However, the four methodologies sug- the argument is that we only know them from descrip- gested in this article are rationalism, empiricism, hi- tions—our own or those of other people (or from per- storicism, and pragmatism, which are well known epi- ceptions, but such perceptions have to be transferred stemological positions, and thus normative principles to descriptions)—and those descriptions or percep- on how to obtain knowledge. It is correct, however, tions will be influenced by culture, goals, interests—in that these positions also rest on different ontological short, subjectivity. In this way, classification is always assumptions. The a priori of rationalism is clear ideas based on properties of objects. Hjørland’s definition of or logical units; for empiricism, it is sense impressi- classification is, however, correctly cited by Wesolek on ons; for historicism, it is change; and, for pragmatism, the same page: “Classification, as defined by Hjørland, the a priori is living and acting in the world. is the ‘sorting of objects based on some criteria se- 4. An anonymous reviewer wrote: “The explanations of lected among the properties of the classified objects.’” the classical theory and prototype theory are not 2. Parrochia and Neuville’s (2013) monograph Towards a strong” and provided a helpful, but long lengthy im- General Theory of Classifications is written from the per- proved description. The choice made here is to try to spective of mathematics but demonstrates surprisingly maintain the short outlines here and later to have the- broad knowledge of classification research, including se theories covered by independent articles. research in the community of knowledge organization 5. A monothetic class is defined in terms of characteri- (see also Parrochia 2016). They, too, find that their stics that are both necessary and sufficient in order to work is about the epistemology of classifications, not a identify members of that class. This way of defining a mathematical textbook or monograph (vii). They write class is also termed the Aristotelian definition of a (xv): class. A polythetic class is defined in terms of a broad set of criteria that are neither necessary nor sufficient. The least we can say is that the field of a mathe- Each member of the category must possess a certain matical theory of classifications is not a com- minimal number of defining characteristics, but none pletely stable domain, and one is led to think, fi- of the features has to be found in each member of nally, that a vast side of it is still to be developed. the category. This way of defining classes is associa- Another reason for the lack of a general theory, ted with Wittgenstein’s concept of “family resem- close to the previous one, is that scientists are blances.” The monothetic type is a type in which all faced with a very difficult problem (finding a members have identical characteristics; the polythetic formalism enough general to apply to any kind type is a type in which all members are similar, but of classifications), for which no complete solu- not identical. tion is known at the present. The distinction between monothetic and polythetic All the same, we think that the research we have classification is discussed by van Rijsbergen (1979, 28- carried on for more than thirty years might be of 29): “An early statement of the distinction between some interest for librarians, logicians, and also for monothetic and polythetic is given by Beckner (1959, scientists in the different fields of empirical sci- 22): ‘A class is ordinarily defined by reference to a set ence, all of whom need to devise their own clas- of properties which are both necessary and sufficient sifications. (by stipulation) for membership in the class. It is possi- But this book has a deeper stake. In fact, pure ble, however, to define a group K in terms of a set mathematics wants also a general theory of clas- G of properties f1, f2, … fn in a different manner. Knowl. Org. 44(2017)No.2 115 Reviews of Concepts in Knowledge Organization

Suppose we have an aggregate of individuals (we shall physics.” The first part of this quotation was elimina- not yet call them a class) such that ted from the same quotation in section 4.1 because it seems confusing and probably wrong that polythetic 1) each one possesses a large (but unspecified) classification is part of formal classification theory number of the properties in G; and has been termed Aristotelian. 2) each f in G is possessed by a large number of 6. The idea is older, however. “The starting point is the these individuals; and, work of the great French botanist Michel Adanson, 3) no f in G is possessed by every individual in who proposed that a member of a class of plants did the aggregate. not need to possess all the defining features of the class, and that a deviant specimen did not need to be The first sentence of Beckner’s statement refers to assigned to a separate class (Adanson 1763, i: cliv the classical Aristotelian definition of a class, which is sqq.) [note omitted]. The important point he made now termed monothetic. The second part defines po- was that creatures should be grouped together on the lythetic. greatest number of features in common, and there is To illustrate the basic distinction, consider the follo- no justification for deciding a priori on the relative im- wing example (Figure 2) of 8 individuals (1-8) and 8 portance of characters in making a natural taxonomy properties (A–H). (Sneath 1962, 292)” (Needham, 1975, 353). 7. An anonymous reviewer wrote: “Just as a mild correc- tion to Lakoff ”s verbal flourishes, the Aristotelian- Classical theory certainly had been the subject of major debate, a debate that had lasted 2000 years and involved some of the finest scholars.” 8. An anonymous reviewer wrote: “I am not sure about Kuhn and Billig in this setting. We are getting drawn off into potentially quite deep cognitive or social psychology. But isn’t our interest storing and retrieving information or knowledge?” Yes, that is our interest, and the claim is that Kuhn and Billig provide important knowledge for this purpose. Kuhn provides the knowledge that we need to consider how different theories or paradigms classify knowledge and relate our decisions to a choice or a negotiation between different views. Billig help us see the problematic assumptions in the cognitive view according to which we have some built-in mechanisms on how to classify knowledge. Figure 2. An illustration of the difference between monothetic 9. “From Cesalpino to Linnaeus, this [downward classifi- and polythetic. cation by logical division] was the almost universally preferred system, particularly in botany (Mayr 1982, The possession of a property is indicated by a plus 158-79). In this methodology, the classifier starts with sign. The individuals 1-4 constitute a polythetic the entire “universe”—let’s say, all animals—and, with group, each individual possessing three out of four of the help of divisional logic, divides them into more and the properties A,B,C,D. The other 4 individuals can more homogeneous groups. A criterion like blood be split into two monothetic classes, {5,6} and {7,8}. temperature, when applied to animals, results in two The distinction between monothetic and polythetic is groups: warm-blooded and cold-blooded animals. By a particularly easy one to make providing the proper- continuing dichotomy, one finally reaches the species ties are of a simple kind, e.g., binary state attributes. level. Even though Linnaeus in his artificial system still When the properties are more complex, the definiti- employed downward classification, it had become evi- ons are rather more difficult to apply, and in any case dent by his time that a reliance on single characters— are rather arbitrary.” and the inevitable arbitrariness of the sequence in Bowker (1998, 256) wrote: “Aristotelian models— which these characters were chosen—could lead to ra- monothetic or polythetic—have traditionally infor- ther artificial systems. At the end of the eighteenth med formal classification theory in a broad range of century, downward classification was therefore replaced sciences, including biological systematics, geology, and by upward classification. In retrospect, it eventually be- 116 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

came evident that the downward procedure actually cal system of another such structure. Apparently produces identification schemes rather than classifica- on the basis of such taxonomic incommensura- tions. It survives today in the form of keys. Curiously, bility, Kuhn asserted a number of antirealist the- in the literature of the philosophy of science, particu- ses about truth, reference and reality. In this pa- larly of logic, classification by logical division has been per, it will be argued, however, that, far from considered the method of classification up to modern leading to antirealist consequences about the rela- times” (Mayr 1995, 420-21). tionship between theory and reality, the taxo- 10. “This system [grouping or upwards classification] most nomic incommensurability thesis may be incor- closely conveys the basic meaning of the term classifi- porated unproblematically within a reasonably cation, that is, to assemble items into classes on the ba- robust scientific realist framework (Sankey 1998, sis of resemblance in observed characters” (Mayr 1995, 7). 421); and “After earlier trials by some herbalists and by Magnol, it was particularly Adanson (1763) who pro- With this theory of kinds, Kuhn redraws the pic- moted classification by grouping. By the first third of ture of scientific revolutions. Since the intercon- the nineteenth century it had become the almost uni- nections among kind terms form a lexical taxon- versal method of classifying plants and animals (Mayr omy, scientific revolutions, which now are limited 1982, 190-208). The empirical rule guiding the taxo- to the meaning change of kind terms, become nomist was well stated by Whewell (1840, v. 1:521): taxonomic changes. A scientific revolution pro- “The maxim by which all systems professing to be na- duces a new lexical taxonomy, in which some tural must be tested is this: that the arrangement obtai- kind terms refer to new referents that overlap ned from one set of characters coincides with the ar- with those denoted by some old kind terms. rangement obtained from another set” (Mayr 1995, Therefore, incommensurability does not result 422). merely from translation failures of individual 11. An anonymous reviewer wrote: “There wants to be concepts. The prerequisite for full translatability emphasis here that the classification is systematic classi- between two taxonomies is not shared features fication [as opposed to conceptual classification].” of individual concepts, but a shared lexical struc- However, the suggested principles are meant to serve ture (Kuhn, 1990b, p. 7). Scientists from rival conceptual classification as well as systematic classifica- paradigms face incommensurability because they tion, and a fundamental view is that conceptual classifi- construct different lexical taxonomies and cation represents the core theory on which systematic thereby classify the world in different ways (Chen classification is based. The reviewer seems to view the 1997, 260; Kuhn 1990b refers to an unpublished two kinds of classification as too dualistic. manuscript). 12. Medin and Aguilar (1999, 104; emphasis in original), for example, wrote: “Why is this notion that categories 14. The following quotation exemplifies the complex pat- are defined by some ‘objective’ similarity controversial? tern of different interests that may be at play behind The main criticism has been that the notion of similari- given classifications—in particular in the domain of ty is too unconstrained to be useful as an explanatory arts: “The work of DiMaggio (1987) has provided the principle (Goodman 1972; Murphy and Medin 1985). theoretical foundation for much research on classifi- Similarity is usually defined in terms of shared proper- cation systems. His concept of artistic classification ties, but Goodman argued that any two things share an systems offers a number of insights. First, the study unlimited number of properties (e.g. robins and of classification systems needs to take into account elephants can move, weight more than two ounces, both the consumption and production of art. Accor- take up space, can be thought about, etc.). Given this ding to DiMaggio (1987: 441), artistic classification apparent flexibility, it may be that we see things as simi- systems consist of ‘the way that the work of artists is lar because they belong to the same category and not divided up both in the heads and habits of consumers vice versa. That is, maybe we can explain similarity in and by the institutions that bound the production and terms of categories.” distribution of separate genres’ [italics in original]. On 13. About incommensurability, see Kuhn (2000), and the one hand, classification systems arise out of pro- consider the following quotations: cesses of social distinction, whereby consumers use cultural objects to mark social boundaries. These ‘rit- Incommensurability arises because it is impossi- ual classifications’ can thus be influenced by social ble to transfer the natural categories employed structural factors at the societal level—such as strati- within one taxonomic structure into the categori- fication systems, elite cohesion, social and geographic Knowl. Org. 44(2017)No.2 117 Reviews of Concepts in Knowledge Organization

mobility, etc.—that generate demand for cultural lism would loom through the still widespread concern boundaries. On the other hand, classification systems for “natural” classifications” (Marradi 1990, 149). are also influenced and mediated by “classification 17. Jon Fjeldså also gave a speech at the University of processes” at the production side. DiMaggio (1987) Copenhagen on 26 March 2014: “Får vi snart en “en- identifies commercial classifications (the classifica- delig” fugleklassifikation?” (Do we soon get a “final” tions used by commercial producers to market their classification of birds?) products), administrative classifications (created and enforced by the state), and professional classifications References (classifications driven by the incentives of artists to differentiate and mark boundaries). The study of Academia das Ciências de Lisboa. 2001. Dicionário da lín- classification systems thus needs to be attentive to the gua portuguesa contemporânea. vol. 1. Lisboa: Verbo. extent to which the categorical demands of consu- Adamson, Robert. 1901. “Classification.” In Dictionary of mers and the categories used by producers overlap, Philosophy and Psychology, ed. James Mark Baldwin. New diverge, or mutually reinforce each other” (van Ven- York, NY: The MacMillan Company, vol. 1: 185. http:// rooij and Schmutz 2015, 799). psychclassics.yorku.ca/Baldwin/Dictionary/defs/C2 15. Nobes and Stadler (2013, 575) examine the degree to defs.htm which classification is determined by who is classify- Adanson, Michel. 1763. Familles des plantes. Paris: Vincent. ing and they refer to Bloor’s (1982, 268) support for Adler, Melissa and Joseph T. Tennis. 2013. “Toward a the claim of Durkheim and Mauss (1903 [2010]) that Taxonomy of Harm in Knowledge Organization Sys- the classification of things reproduces a pattern of tems.” Knowledge 0rganization 40: 266-72. social arrangements more than a pattern of the Andersen, Hanne, Peter Barker and Xiang Chen. 1996. things. Durkheim and Mauss found that our classiﬁ- “Kuhn’s Mature Philosophy of Science and Cognitive cations necessarily are hierarchical because human so- Psychology.” Philosophical Psychology 9: 347-63. cial organization, from which our view of nature ul- Andersen, Hanne, Peter Barker and Xiang Chen. 2006. timately ﬂows, is itself hierarchical. This view has The Cognitive Structure of Scientific Revolutions. Cambridge, been strongly criticized as “sociologism,” but is, as we UK: Cambridge University Press. saw, defended by Bloor (1982). Aristotle. (350 BCE) 2006. Topics trans. William Adair 16. Marradi wrote: “The opposition between ‘natural’ and Pickard-Cambridge). Overland Park, KS: Digireads ‘artificial’ classification is a recurring theme in the last Publishing. http://classics.mit.edu/Aristotle/topics.1. two centuries. Cohen and Nagel have cogently argued i.html that ‘any division ... according to some actual trait ar- Bailey, Kenneth D. 1994. Typologies and Taxonomies: An In- bitrarily chosen is perfectly natural ... [but it] may al- troduction to Classification Techniques. London: Sage Publi- so be said to be artificial, in the sense that we select cations. the trait’ (1934, 223). Yet, many scholars have conti- Bather, Francis Arthur. 1927. “Biological Classifications: nued stressing ‘naturality’ as a desirable property, by Past and Future.” Quarterly Journal of the Geological Society gradually re-interpreting it in terms of significant rela- 83: 63-104. tionships with other classifications (Hempel 1961, Beckner, Morton. 1959. The Biological Way of Thought. New version reprinted 1965, 146-7; Kaplan 1964, 50), utili- York, NY: Columbia University Press. ty ‘for a wider range of inductive generalizations’ Beghtol, Clare. 2003. “Classification for Information Re- (Gilmour 1940, 466), ‘systematic import’ (Huxley trieval and Classification for Knowledge Discovery: 1940; Hempel 1952; Sandri 1969), links with theory Relationships between “Professional” and “Naïve” (Hempel 1952; Bunge 1967, 83), ‘projectibility of dis- Classifications.” Knowledge Organization 30: 64-73. criminating concepts’” (Sandri 1969, 99 ff.). In short, Beghtol, Clare. 2004. “Exploring New Approaches to the the concept of “natural classification” has been trans- Organization of Knowledge: The Subject Classifica- ferred from the ontological to the epistemological tion of James Duff Brown.” Library Trends 52, no. 4: domain. However, as Tiryakian (1968, 177) has re- 702-18. marked, “the reification of typologies is a frequent Beghtol, Clare. 2010 “Classification Theory.” In Encyclope- temptation and pitfall.” In a typical manual of the dia of Library and Information Sciences, 3rd ed., eds. Marcia neopositivist period one can still read a statement as J. Bates and Mary Niles Maack. Boca Raton, FL: CRC plain as “A natural class is based on the fundamental Press, 2: 1045-60. character of things” (Lenzel 1938, 32). And one may Biggam, Carole Patricia. 2015. The Semantics of Colour: A suspect that, if the epistemological coat of paint were Historical Approach. Cambridge, UK: Cambridge Uni- scraped off, quite substantial traces of rusty essentia- versity Press. 118 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

Billig, Michael. 1996. Arguing and Thinking. A Rhetorical Cranz, David. 1767. The History of Greenland: Containing a Approach to Social Psychology, 2nd ed. Cambridge, UK: Description of the Country and its Inhabitants. Vol. I-II. Cambridge University Press. London: The Brethren’s Society for the Furtherance Bliss, Henry E. 1929. The Organization of Knowledge and the of the Gospel among the Heathen. System of the Sciences. New York, NY: Henry Holt and Cutter, Charles Ammi. 1876. Rules for a Printed Dictionary Company. Catalogue. Washington: Government Printing Office. Bliss, Henry Evelyn. 1933. The Organization of Knowledge in Dahlberg, Ingetraut. 2010. “International Society for Libraries and the Subject-approach to Books. New York, Knowledge Organization (ISKO).” Encyclopedia of li- NY: H. W. Wilson. brary and information sciences, 3rd ed., ed.vMarcia J. Bates Bliss, Henry E. 1935. A System of Bibliographical Classifica- and Mary Niles Maack. Boca Raton, FL: CRC Press, 4: tion. New York: The H. W. Wilson Company. 2941-49. Bloor, David. 1982. “Durkheim and Mauss revisited: Darwin, Charles. 1859. On the Origin of Species By means of Classification and the Sociology of Knowledge.” Stud- Natural Selection, or The Preservation of Favoured Races in ies in History and Philosophy of Science 13: 267-97. the Struggle for Life. London: J. Murray. Bowker, Geoffrey C. 1998. “The Kindness of Strangers: DiMaggio, Paul. 1987. Classification in Art. American So- Kinds and Politics in Classification Systems.” Library ciological Review 52: 440-55. Trends 47, no. 2: 255-92. Dousa, Thomas M. 2009. “Evolutionary Order in the Clas- Bowker, Geoffrey C. and Susan Leigh Star. 1999. Sorting sification Theories of C. A. Cutter and E. C. Richard- Things Out: Classification and its Consequences. Cambridge, son: Its Nature and Limits.” In Proceedings of the 2009 MA: MIT Press. North American Symposium on Knowledge Organization, June Bunge, Mario A. 1967. Scientific Research, vol. I: The Search 18-19, 2009, Syracuse, New York: 76-90 https://jour for System. New York, NY: Springer. nals.lib.washington.edu/index.php/nasko/article/ Campbell, Joseph Keim, Michael O’Rourke and Matthew download/12810/11292 H. Slater. eds. 2012. Carving Nature at its Joints: Natural Drucker, Donna J. 2014. The Classification of Sex. Alfred Kinds in Metaphysics and Science. Cambridge, MA: MIT Kinsey and the Organization of Knowledge. Pittsburgh, PA: Press. University of Pittsburgh Press. Candolle, Augustin-Pyramus de. 1813. Théorie élémentaire de Dupré, John. 1993. The Disorder of Things. Metaphysical la Botanique ou exposition des principes de la classification na- Foundations for the Disunity of Science. Cambridge, MA: turelle et de l’art d’écrire et d’étudier les végétaux. Paris: De- Harvard University Press. terville. Dupré, John. 2002. Humans and Other Animals. Oxford, Capecchi, Vittorio and Frank Möller. 1968. “Some Appli- UK: Oxford University Press. cations of Entropy to the Problems of Classification.” Dupré, John. 2006. “Scientific Classification.” Theory, Cul- Quality & Quantity 2: 63-84. ture & Society 23, nos. 2-3: 30-32. Carpenter, William Benjamin. 1847. Zoology: A Systematic Durkheim, Émile and Marcel Mauss. 1903. “De quelques Account of the General Structure, Habits, Instincts, and Uses formes primitives de classification: Contribution à of the Principal Families of the Animal Kingdom. London: l’étude des représentations collectives.” Année Sociologi- W. S. O r r. que VI (1901-1902): 1-72. Chakravartty, Anjan. 2011. Scientific Realism and Onto- Durkheim, Émile and Marcel Mauss. 2010. Primitive Clas- logical Relativity. The Monist 94: 157-80. sification, trans. from the French and with an introduc- Chan, Lois Mai. 1994. Cataloging and Classification: An In- tion by Rodney Needham. London: Routledge. troduction, 2nd ed. New York, NY: McGraw-Hill. Eiser, J. Richard. 1980. Cognitive Social Psychology. London: Chen, Xiang. 1997. Thomas Kuhn”s Latest Notion of In- McGraw Hill. commensurability. Journal for General Philosophy of Sci- Faria, Maria Isabel and Maria da Graça Pericão. 2008. Di- ence/Zeitschrift für allgemeine Wissenschaftstheorie 28: 257-73. cionário do livro: Da escrita ao livro electrónico. Coimbra: Cohen, Morris Raphael and Ernst Nagel. 1934. An Intro- Almedina. duction to Logic and Scientific Method. New York, NY: Farradane, Jason E. L. 1950. “A Scientific Theory of Harcourt. Classification and Indexing and its Practical Applica- Collin, Finn. 1993. Social Constructivism without Para- tions.” Journal of Documentation 6: 83-99. dox. Danish Yearbook of Philosophy 28: 24-46. Farradane, Jason E. L. 1952. “A Scientific Theory of Condorcet, Marie Jean Antoine Nicolas de Caritat, Mar- Classification and Indexing: Further Considerations.” quis de. 1777. “Sur les familles naturelles des plantes, Journal of Documentation 8: 73-92. et en particulier sur celle des renoncules.” Histoire de Feger, Hubert. 2001. “Classification: Conceptions in the l’Académie Royale des Sciences (Paris) Année 1773: 34-36. Social Sciences.” In International Encyclopedia of the Social Knowl. Org. 44(2017)No.2 119 Reviews of Concepts in Knowledge Organization

and Behavioral Sciences, ed. Neil J. Smelser and Paul B. Henry, Devin. 2011. “Aristotle’s Pluralistic Realism.” The Baltes. Amsterdam: Elsevier, 3: 1966-73. Monist 94: 197-220. Feger, Hubert. 2015. “Classification: Understandings of Hjørland, Birger. 2003. “Fundamentals of Knowledge in the Social Sciences.” In International Encyclopedia of Organization.” Knowledge Organization 30: 87-111. the Social and Behavioral Sciences 2nd ed., ed. James D. Hjørland, Birger. 2009. “Concept Theory.” Journal of the Wright. Amsterdam: Elsevier, 3: 805-10. American Society for Information Science and Technology 60: Fjeldså, Jon. 2013. Avian Classification in Flux. In Hand- 1519-36. book of the Birds of the World, ed. Josep del Hoyo, An- Hjørland, Birger. 2011. “The Importance of Theories Of drew Elliott, Jordi Sargatal and David A. Christie. Spe- knowledge: Indexing and Information Retrieval as an cial volume 17: 77-146. Barcelona: Lynx Editions. Example.” Journal of the American Society for Information Fox, Melodie J. 2011. “Prototype Theory: An Alternative Science and Technology 62: 72-77. Concept Theory for Sex and Gender?” Knowledge Hjørland, Birger. 2013. “Citation Analysis: A Social and Organization 38:328-34. Dynamic Approach to Knowledge Organization.” In- Fox, Melodie J. and Hope A. Olson. 2012. “Feminist formation Processing & Management 49: 1313-25. Epistemologies and Knowledge Organization.” In Cul- Hjørland, Birger. 2014. “Is Facet Analysis based on Ra- tural Frames of Knowledge, ed. Richard P. Smiraglia and tionalism? A Discussion of Satija (1992), Tennis (2008), Hur-li Lee. Würzburg: Ergon Verlag, 79-98. Herre (2013), Mazzocchi (2013b), and Dousa and Frické, Martin. 2012. Logic and the Organization of Informa- Ibekwe-SanJuan (2014).” Knowledge Organization 41: 369- tion. New York, NY: Springer. 76. Frické, Martin. 2016. “Logical Division.” Knowledge Or- Hjørland, Birger. 2015. “Are Relations in Thesauri ‘Context ganization 43: 539-49. A version is also available as Free, Definitional, and True in all Possible Worlds’?” http://www.isko.org/cyclo/logical_division Journal of the Association for Information Science and Technol- Galison, Peter and David J. Stump eds. 1996. The Disunity ogy 66: 1367-73. of Science: Boundaries, Contexts, and Power. Stanford, CA: Hjørland, Birger. 2016a. “Informetrics needs a Founda- Stanford University Press. tion in the Theory of Science.” In Theories of Infor- Gilmour, John Scott Lennox. 1940. “Taxonomy and Phi- metrics and Scholarly Communication, ed. Cassidy Sugi- losophy,” In The New Systematics, ed. Julian Huxley. Ox- moto. Berlin: Walter de Gruyter, 20-46. ford, UK: Clarendon Press, 461-74. Hjørland, Birger. 2016b. “The Paradox of Atheoretical Gitelman, Lisa ed. 2013. “RAW data” is an Oxymoron. Classification.” Knowledge Organization 43: 313-23. Cambridge, MA: MIT Press. Hjørland, Birger. 2017. “Subject (Of Documents).” Gnoli, Claudio. 2006. “Phylogenetic Classification.” Knowledge Organization 44: 55-64. Knowledge Organization 33: 138-52. Hjørland, Birger and Hanne Albrechtsen. 1999. “An Goodman, Nelson. 1972. Seven Strictures on Similarity. Analysis of Some Trends in Classification Research.” In Problems and Projects, ed. Nelson Goodman. New Knowledge Organization 27: 131-39. York, NY: Bobbs-Merrill, 437-46. Hjørland, Birger and Karsten Nissen Pedersen. 2005. “A Grove, Andrew. 2010. “Taxonomy.” In Encyclopedia of Li- Substantive Theory of Classification for Information brary and Information Sciences, 3rd ed., ed. Marcia J. Bates Retrieval.” Journal of Documentation 61: 582-97. and Mary Niles Maack. Boca Raton, FL: CRC Press, 7: Hudon, Michéle, Sabine Mas and Dominique Gazo. 2005. 5139-48. “Structure, Logic, and Semantics in Ad Hoc Classifica- Hedden, Heather. 2016. Accidental Taxonomist. Second edition Schemes Applied to Web-based Libraries in the tion. Medford, NJ: Information Today. Field of Education. Canadian Journal of Information and Hempel, Carl Gustav. 1952. Fundamentals of Concept Forma- Library Science 29: 265-88. tion in Empirical Science. Chicago, IL: Chicago University Hull, David L. 1998. “Taxonomy.” In Routledge Encyclopedia Press. of Philosophy, ed. Edward Craig. London: Routledge 9: Hempel, Carl Gustav. 1961. “Fundamentals of Taxon- 272-76. omy.” In Field Studies in Mental Disorders, ed. Joseph Huxley, Julian ed. 1940. The New Systematics. Oxford: Clar- Zubin. New York, NY: Grune & Stratton. (Quotations endon Press. from the version are reprinted in: Hempel, Carl G. Huxley, Thomas Henry. 1869. AN introduction to the Classifi- 1965. Aspects of Scientific Explanation and other Essays in the cation of Animals. London: John Churchill & Sons. Philosophy of Science. Glencoe, IL: Free Press, 137-54). https://archive.org/stream/anintroductiont00huxlgoog Hennig, Willi. 1966. Phylogenetic Systematics, trans. D. Dwight #page/n4/mode/2up Davis and Rainer Zangerl. Urbana, IL: University of Il- ISO 5127-6. 1988. “Documentation et information— linois Press. Vocabulaire: Partie 6: Langages documentaires.” In 120 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

Documentation et information: Recueil de normes ISO I. Mai, Jens Erik. 2004. “Classification in Context: Relativ- Genève: ISO, 87-111. ity, Reality, and Representation.” Knowledge Organization Jacob, Elin K. 2004. “Classification and Categorization: A 31: 39-48. Difference that Makes a Difference.” Library Trends 52, Mai, Jens Erik. 2011. “The Modernity of Classification.” no. 3: 515-40. Journal of Documentation 67: 710-30. Jung, Carl G. 1971. The Collected Works of C. G. Jung. Vol- Margolis, Eric. 1994. “A Reassessment of the Shift from ume 6; Psychological Types. Princeton: Princeton Univer- the Classical Theory of Concepts to Prototype The- sity Press. ory.” Cognition 51: 73-89. Kaplan, Abraham. 1964. The Conduct of Inquiry. San Fran- Marradi, Alberto. 1990. “Classification, Typology, Taxon- cisco, CA: Chandler. omy.” Quality and Quantity 24, no. 2: 129-57. Prepublica- Kedrow, Bonifatii Mikhailovich. 1975. Klassifizierung der tion [not identical] available at: http://web. Wissenschaften 1-2. Moskau: Verlag Progress. archive.org/web/20040705070709/http://www.unibo. Khalidi, Muhammad Ali. 2013. Natural Categories and Hu- edu.ar/marradi/classqq.pdf man Kinds: Classification in the Natural and Social Sciences. May, Ronald W. 1982. “Discriminant Analysis in Cluster Cambridge, UK: Cambridge University Press. Analysis.” In Classifying Social Data, ed. Herschel C. Klaus, Georg. 1976. “Klassifikation.” In Philosophisches Hudson. San Francisco, CA: Jossey-Bass, 39-55. Wörterbuch Band 1-2. 12. Aufl, ed. Georg Klaus and Mayr, Ernst. 1982. The Growth of Biological Thought. Cam- Manfred Buhr. Leipzig: Bibliographisches Institut, vol. bridge, MA: Harvard University Press. 1: 628-29. Mayr, Ernst. 1995. “Systems of Ordering Data.” Biology Kuhn, Thomas S. 1962. The structure of scientific revolutions. and Philosophy 10: 419-34. Chicago, IL: University of Chicago Press. Mayr, Ernst and Walter Joseph Bock. 2002. “Classifica- Kuhn, Thomas S. 1974. “Second Thoughts on Para- tions and Other Ordering Systems.” Journal of Zoological digms.” In The Structure of Scientific Theories, ed. Freder- Systematics and Evolutionary Research 40, no. 4: 169-94. ick Suppe. Urbana, IL: University of Illinois Press, Mazzocchi, Fulvio. 2013. “Images of Thought and their 459-82. (Reprinted in Thomas S. Kuhn. 1977. The Es- Relation to Classification: The Tree and the Net.” sential Tension. Chicago, IL: University of Chicago Knowledge Organization 40: 366-74. Press, 293-319). Mazzocchi, Fulvio. 2017. “Relations in KOS: Is it Possi- Kuhn, Thomas S. 2000. The Road Since Structure: Philosophi- ble to Couple a Common Nature with Different cal Essays, 1970–1993. Chicago, IL: University of Chi- Roles?” Journal of Documentation 73: 368-83. cago Press. McKenna, Malcolm C. and Susan K. Bell. 1998. Classifica- Lakoff, George. 1987. Women, Fire and Dangerous Things: tion of Mammals Above the Species Level. New York, NY: What Categories Reveal about the Mind. Chicago, IL: Uni- Columbia University Press. versity of Chicago Press. Medin, Douglas L. and Cynthia Aguilar. 1999. “Categori- Lancaster, F. Wilfrid. 1998. Indexing and Abstracting in The- zation.” In The MIT Encyclopedia of the Cognitive Sciences, ory and Practice, 2nd ed. London: Library Association. ed. Robert A. Wilson and Frank C. Keil. Cambridge, Landweh, Carl E., Alan R. Bull, John P. McDermott and MA: The MIT Press, 104-6. William S. Choi. 1994. “A Taxonomy of Computer Meinhardt, Helmuth et al. 1984. “Ordnung.” In Historisches Program Security Flaws.” ACM Computing Surveys 26: Wörterbuch der Philosophie, ed. Joachim Ritter and Karl- 211-54. fried Gründer. Darmstadt: Wissenschaftliche Buchge- Lenzel, Victor F. 1938. “Procedures of Empirical Sci- sellschaft, Band 6: 1249-1310. (Meinhardt, Helmut, ence.” In International Encyclopedia of Unified Science, ed. “Ordnung I Antike,” 1251-54; Hübener, Wolfgang, Otto Neurath, Rudolf Carnap and Charles W. Morris. “Ordnung II Mittelalter,” 1254-79; Dierse, Ulrich, Chicago, IL: University of Chicago Press, vol. 1, no. 5. “Ordnung III Neuzeit,” 1280-1303; Steiner, Hans- Linnaeus, Carl. 1767. Fundamenta entomologiæ. Uppsala: John Georg, “Ordnung IV Mathematik Logik, ” 1303-9; Sau- Edman. https://web.archive.org/web/ 200707132 er, Werner, “Ordnung der Begriffe,” 1309-10). 35756/http://huntbot.andrew.cmu.edu/HIBD-PDF/ Miksa, Francis L. 1994. “Classification.” In Encyclopedia of LinnaeanDiss/Liden-154.pdf Library History, ed. Wayne A. Wiegand and Donald G. McGarry, Kevin. 1991. “Epilogue: Differing Views of Davis. New York, NY: Garland Publishing, 144-53. Knowledge.” In Knowledge and Communication. Essays on Miksa, Francis L. 1998. The DDC, the Universe of Knowledge, the Information Chain, ed. Arthur Jack Meadows. Lon- and the Post-modern Library. Albany, NY: Forest Press. don: Library Association, 132-52. Mill, John Stuart. 1843. A System of Logic, ratiocinative and Inductive: Being a Connected View of the Principles of Evi- Knowl. Org. 44(2017)No.2 121 Reviews of Concepts in Knowledge Organization

dence and the Methods of Scientific Investigation. London: Ranganathan, Shiyali Ramamrita. 1967. Prolegomena to Li- John W. Parker, West Strand. brary Classification, 3rd ed. London: Asia Publishing Mill, John Stuart. 1872. A System of Logic, Ratiocinative and House. Inductive: Being a Connected View of the Principles of Evi- Reid, Thomas. 1785. “Abstraction.” Reprinted in James F. dence and the Methods of Scientific Investigation vol. 1-2, 8th Bennett, ed. 1855. Essays on the Intellectual Powers of Man ed.. London: Longmans, Green, Reader, and Dyer. by Thomas Reid, 6th ed. (redacted text). Boston, MA: Moss, Wilfred Raymond. 1964. “Categories and Rela- Phillips, Sampson, and Company, 298-330. https:// tions. Origins of Two Classification Theories.” Ameri- babel.hathitrust.org/cgi/pt?id=hvd.hwaili;view=1up; can Documentation 15: 296-301. seq=17 Müller-Wille, Staffan. 2007. “Collection and Collation: Richardson, Ernest Cushing. 1901. Classification, Theoretical Theory and Practice of Linnaean Botany.” Studies in and Practical: Vol. I: The Order of the Sciences; vol. II: The History and Philosophy of Biological and Biomedical Sciences Classification of Books. New York, NY: Charles Scrib- 38: 541-62. ner’s Sons. Müller-Wille, Staffan. 2013. “Systems and How Linnaeus Ritter, Joacim, Karlfried Gründer and Gabriel Gottfried, Looked at them in Retrospect.” Annals of Science 3: eds. 1971-2007. Historisches Wörterbuch der Philosophie Bd. 305-17. 1-13. Basel: Schwabe & Co. Murphy, Gregory L. and Douglas L. Medin. 1985. “The Rosch, Eleanor. 1978. “Principles of Categorization.” In Role of Theories in Conceptual Coherence.” Psychologi- Cognition and Categorization, ed. Eleanor Rosch and Bar- cal Review 92: 289-316. bara Bloom Lloyd eds. Hillsdale, NJ: L. Erlbaum As- Needham, Rodney. 1975. “Polythetic Classification: Con- sociates, 27-48. vergence and Consequences.” Man. New Series 10: 349- Rysiew, Patrick. 2016. "Naturalism in Epistemology.” In The 69. Stanford Encyclopedia of Philosophy. Spring 2016 ed., ed. Nobes, Christopher and Christian Stadler. 2013. How Ar- Edward N. Zalta. http://plato.stanford.edu/archives/ bitrary are International Accounting Classifications? spr2016/entries/epistemology-naturalized/ Lessons from Centuries of Classifying in Many Disci- Sandri, Giorgio. 1969. “On the Logic of Classification,” plines, and Experiments with IFRS Data. Accounting Or- Quality & Quantity 3: 80-124. ganizations and Society 38: 573-95. Sankey, Howard. 1998. “Taxonomic Incommensurability.” OED Online. 1990. Oxford English Dictionary: The Definitive International Studies in the Philosophy of Science 12: 7-16. Record of the English Language. Oxford, UK: Oxford Satija, Mohinder P. 1992. Book review of Knowledge and University Press. Communication: Essays on the Information Chain, ed. by A. Padian, Kevin. 1999. “Charles Darwin’s Views of Classi- J. Meadows. International Classification 19: 39-41. fication in Theory and Practice.” Systematic Biology 48: Satija, Mohinder P. 2000. “Library Classification: An Es- 352-64. say in Terminology.” Knowledge Organization 27: 221-9. Pando, Daniel Abraao and Carlos Candido de Almeida. Schmidt, Kjeld and Ina Wagner. 2004. “Ordering Sys- 2016. “Knowledge Organization in the Context of tems. Coordinative Practices and Artifacts in Architec- Postmodernity from the Theory of Classification Per- tural Design and Planning.” The Journal of Collaborative spective.” Knowledge Organization 43: 113-17. Computing 13: 349-408. Parrochia, Daniel 2016. “Classification.” In The Internet Shera, Jesse H. 1965. Libraries and the Organization of Encyclopedia of Philosophy, ed. James Fieser and Bradley Knowledge. Hamden, CT: Archon Books. Dowden. Martin, TN: University of Tennessee at Mar- Simões, Maria da Graça, M. Cristina V. de Freitas and tin. http://www.iep.utm.edu/classifi/ Blanca Rodríguez-Bravo. 2016. “Theory of Classifica- Parrochia, Daniel and Pierre Neuville. 2013. Towards a tion and Classification in Libraries and Archives: Con- General Theory of Classifications. Bäsel: Birkhaüser. vergences and Divergences.” Knowledge Organization 43: Parry, William T. and Edward A. Hacker. 1991. Aristotelian 530-8. Logic. New York, NY: State University of New York Slavic, Aida. 2000. “A Definition of Thesauri and Classi- Press. fication as Indexing Tools.” DCMI note. http:// Pihlström, Sami. 2009. Pragmatist Metaphysics. An Essay on the dublincore.org/documents/thesauri-definition/ Ethical Grounds of Ontology. New York, NY: Continuum. Sloan, Phillip R. 1981. “Classification.” In Dictionary of the Plato. (c. 370 BC) 1995. Phaedrus, trans. with introduction History of Science, ed. William F. Bynum, E. Janet and notes by Alexander Nehamas and Paul Woodruff. Browne and Roy Porter. Princeton, NJ: Princeton Cambridge, MA: Hackett Publishing Co, Inc. University Press, 68-71. Popper, Karl Raimund. 1959. The Logic of Scientific Discov- Sneath, Peter Henry Andrews. 1962. “The Construction ery. London: Hutchinson. of Taxonomic Groups.” In Microbial Classification, ed. 122 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

Geoffrey Clough Ainsworth and Peter Henry An- tion of OAI Metadata using the DDC Taxonomy.” In drews Sneath. Cambridge, UK: Cambridge University Advanced Language Technologies for Digital Libraries. Inter- Press, 289-332. national Workshops on NLP4DL 2009, Viareggio, It- Soergel, Dagobert. 2004. “Information Organization.” In aly, June 15, 2009 and AT4DL 2009, Trento, Italy, Sep- Berkshire Encyclopedia of Human-Computer Interaction, ed. tember 8, 2009, ed. Raffaella Bernardi, Frederique William Sims Bainbridge. Great Barrington, MA: Segond and Ilya Zaihrayeu. Lecture Notes in Com- Berkshire Publishing Group LLC, 1: 355-60. puter Science v6699. Heidelberg: Springer, 29-40. Sokal, Robert R. and Peter H. A. Sneath. 1963. Principles Wesolek, Andrew. 2012. “Wittgensteinian Support for of Numeric Taxonomy. San Francisco, CA: W.H. Free- Domain Analysis in Classification.” Library Philosophy man. and Practice paper 795. http://digitalcommons.unl.edu/ Spiteri, Louise F. 2008. “Concept Theory and the Role of libphilprac/795 Conceptual Coherence in Assessments of Similarity.” Whewell, William. 1840. The Philosophy of the Inductive Sci- Proceedings of the American Society for Information Science ences: Founded upon their History. London: J.W. Parker. and Technology 45: 1-12. Wilkins, John S. 2013. “Biological Essentialism and the Stevens, Peter F. 1998. “Linnaeus, Carl von (1707–78).” In Tidal Change of Natural Kinds.” Science and Education Routledge Encyclopedia of Philosophy, version 1.0, ed. Ed- 22: 221-40. http://philpapers.org/rec/WILBEA ward Craig. London: Routledge. doi:10.4324/97804 Wittgenstein, Ludwig. 1953. Philosophical Investigations. Ox- 15249126-Q059-1 ford, UK: Basil Blackwell. Stevens, Peter F. 2016. “Schools of Classification.” In En- WordNet 3.1. “Classification.” http://wordnetweb.princeton. cyclopedia of Evolutionary Biology, ed. Richard M. Kliman. edu/perl/webwn?s=classification San Diego, CA: California Academic Press, 3: 494-8. Suppe, Frederick. 1989. “Classification.” In International Appendix: Encyclopedia of Communications, ed. Erik Barnouw. Ox- A Sample of Definitions of Classification ford, UK: Oxford University Press, 1: 292-96. (Chronological) Sutcliffe, John Philip. 1993. “Concept, Class, and Cate- gory in the Tradition of Aristotle.” In Categories and The intention is to provide a comprehensive list of defi- Concepts, ed. Iven van Mechelen, James Hampton, nitions of classification, and the idea is to update the list Ryszard S. Michalski and Peter Theuns. London: Aca- when new definitions are discovered in the literature. The demic Press, 35-65. sources for such definitions are surprisingly few and Svenonius, Elaine. 2000. The Intellectual Foundation of Infor- meager. The Oxford English Dictionary provides a number mation Organization. Cambridge, MA: The MIT Press. of quotations back to 1767 but misses important ones Svenonius, Elaine. 2004. “The Epistemological Founda- such as Darwin (1859) and Huxley (1869) (see also below tions of Knowledge Representations.” Library Trends under 2010). Historisches Wörterbuch der Philosophie Bd. 1-13 52, no. 3: 571-87. (edited by Joachim Ritter) has no article “Klassifikation” Taylor, Arlene G. 1999. The Organization of Information. or “Taxonomie” (but does have one, for example, “Kate- Englewood, CO: Libraries Unlimited, Inc. gorie, Kategorienlehre,” “Ordnung” and “System, Sys- Tiryakian, Edward A. 1968. “Typologies.” In International tematik, Systematisch”). McKenna and Bell (1998, 11-33) Encyclopedia of the Social Sciences, ed. David L. Sills. New provide an overview of the history and theory of classifi- York, NY: Macmillan, 16: 177-85. cation and state: “The word “classification” was not part van Rijsbergen, Cornelis Joost. 1979. Information Retrieval. of the scientific literature until the last decades of the Second edition. London: Butterworths. eighteenth century. The first use of which we are aware van Venrooij, Alex and Vaughn Schmutz. 2015. “Classifi- occurs in a botanical paper by the Marquis de Condorcet cations in Popular Music.” In International Encyclopedia (1777: 35).” However, earlier uses are listed below. of the Social and Behavioral Science, 2nd ed. James D. Wright. Amsterdam: Elsevier, 3: 799-804. Plato c.370 BC Veloso, Adriano and Wagner Meira Jr. 2011. Demand-driven Socrates Associative Classification. London: Springer. That of dividing things again by classes, where the Venter, J. Craig, Karin Remington, John F. Heidelberg, natural joints are, and not trying to break any part, af- Aaron L. Halpern, et al. 2004. “Environmental Ge- ter the manner of a bad carver. (Plato c.370 BC, nome Shotgun Sequencing of the Sargasso Sea.” Sci- Phaedrus 265e) ence 304 no. 5667: 66-74. Waltinger, Ulli, Alexander Mehler, Mathias Lösch and Wolfram Horstmann. 2011. “Hierarchical Classifica- Knowl. Org. 44(2017)No.2 123 Reviews of Concepts in Knowledge Organization

Aristotle ([350 BCE] 2006) William Whewell (1840, 1:xxxiii, XCV, emphasis original) A “genus” is what is predicated in the category of es- The attempts at Natural Classification are of three sence of a number of things exhibiting differences in sorts; according as they are made by the process of kind. We should treat as predicates in the category of blind trial, of general comparison, or of subordination of essence all such things as it would be appropriate to characters. The process of Blind Trial professes to make mention in reply to the question, “What is the object its classes by attention to all the characters, but with- before you?”; as, for example, in the case of man, if out proceeding methodically. The process of General asked that question, it is appropriate to say “He is an Comparison professes to enumerate all the characters, animal” (Book I, chap. 1, part 5). and forms its classes by the majority. Neither of these Note (Parrochia 2016; see also Sutcliffe 1993): methods can really be carried into effect. The method The logic of classifications, which remains, in this of Subordination of Characters considers some char- time, the Aristotelian logic, receives practically no new acters as more important than others; and this method development until the 18th century. gives more consistent results than the others. This method, however, does not depend upon the Idea of Michel Adanson (1763, clvi) Likeness only, but introduces the Idea of Organization All parts and qualities, or properties and faculties of or Function. plants ... barring not even one” ought to be considered before attempting a classification. John Stuart Mill (1843, 2.4.7:299-300) Classification, thus regarded, is a contrivance for the David Cranz (1767, 1:ix) best possible ordering of the ideas of objects in our I have described what belongs to this science, not ac- minds; for causing the ideas to accompany or succeed cording to the classifications [Ger. Eintheilungen] and one another in such a way as shall give us the greatest characteristics, which are generally adopted by, and command over our knowledge already acquired, and needful for the modern naturalists, but according to a lead more directly to the acquisition of more. The gen- certain affinity or likeness. eral problem of Classification, in reference to these purposes, may be stated as follows: To provide that Carl Linnaeus (1767, 152 majuscule in original) things shall be thought of in such groups, and those Natura Insectorum per plures eorum ætates jam per- groups in such an order, as will best conduce to the re- specta, superest ut systematice eadem contemplemur. membrance and to the ascertainment of their laws. Recta autem eorum CLASSIFICATIO vitam huic sci- entiæ & facultatem conciliat, ubi singula insecta suum William Benjamin Carpenter (1847, I. §2) quasi nomen ipsa produnt. The object of all Classification ... [is] to bring together [The nature of insects through their several stages of those beings which most resemble each other and to life, having already been examined, it remains to con- separate those that differ. template them systematically. Now the correct CLAS- SIFICATION of them [sci., insects] furnishes life and Charles Darwin (1859, 420) means to this science, where individual insects them- all true classiﬁcation is genealogical …. selves produce, as it were, their own name.—trans. by (See also Mayr and Bock 2002; Padian 1999.) Thomas Dousa.] Thomas Henry Huxley (1869, 1) Thomas Reid (1785, 191) By the classification of any series of objects is meant Our ability to distinguish and give names to the differ- the actual, or ideal, arrangement together of those ent attributes belonging to a single thing goes along which are like and the separation of those which are with an ability to observe that many things have cer- unlike, the purpose of this arrangement being to facili- tain attributes in common while they differ in others. tate the operation of the mind in clearly conceiving This enables us to put the countless hordes of indi- and retaining in the memory, the characters of the ob- viduals into a limited number of classes, which are jects in question. called “kinds” and “sorts”—and in the scholastic language called “genera” and “species” (here quoted from Charles Ammi Cutter (1876, 10) Frické 2012, 25). Class, a collection of objects having characteristics in common.

124 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

Robert Adamson (1901, 1:185) Jason E. L. Farradane (1950, 87). The process of arranging the objects of some prov- A classification indicates the relations between items ince of experience into kinds or groups, characterized of knowledge. by the possession of common marks. As ordinarily defined, it involves more than logical Jason E. L. Farradane (1952, 73-74). DIVISION (q.v.), the rules of which furnish the A classification represents a theory of the structure of minimal conditions of the process. In addition, classi- knowledge, i.e. of the relations between the different fication takes into account (1) either the specific pur- parts of knowledge. The basic problem is to deter- pose of the arrangement, or (2) the natural conjunc- mine what these relations are and how they link the tions of marks which are of most importance. In ei- different concepts from our knowledge into a coher- ther case, the aim of classification is to render possible ent structure. The arbitrary or “deductive” subdivision the greatest number of general propositions regarding of an assumed total of knowledge cannot give a true the objects, and so to facilitate the complete and sys- representation of these relations, which do not consist tematic survey of them. The ideal of a classification only of groupings of a class and its members, or divi- that is not determined by special, human ends, as e.g. sion of a whole into its parts. It was shown that a clas- in classification of occupations in a census return, is to sification must be constructed “inductively,” or up- copy in its systematic arrangement the real order of wards, piecing together known fragments of relations. interdependence in the things themselves. What is called “artificial,” as opposed to natural classification, Jesse H. Shera (1965, 120) differs in degree only, not in kind. Literature: MILL, Classification is the crystallization or formalizing of in- Logic, Bk. IV. chaps. vii, viii; VENN, Empirical Logic, ferential thinking, born of sensory perception, condi- chap. xxx; JEVONS, Princ. of Sci., chap. xxx. (R.A.). tioned by the operation of the human brain, and shaped by human experience. It lies at the foundation of all Ernest Cushing Richardson (1901, 1) thought, but it is pragmatic and it is instrumental. Classification is the “putting together of like things, or more fully described, it is the arranging of things ac- Jesse H. Shera (1965, 127) cording to likeness and unlikeness. It may also be ex- He [the librarian] must appreciate classification, not as pressed as the sorting and grouping of things. It is con- a tool, but as a discipline in which is to be studied the venient sometimes, to speak of “likeness and unlike- reaction and response of a living mind to the record ness” but really in classification it is “likeness” which left by a distant and usually unknown mind; a disci- rules while “unlikeness” is merely what is left over when pline that seeks to achieve a better understanding of likeness has been defined. The “putting together of like the changing patterns of thought and the points of things” is therefore the fullest and most exact form of contact at which they can be related to specific units the definition. of recorded information.

Henry E. Bliss (1935, 3) (Drucker 2014 is, however, a work demonstrating that In dealing with the multiplicity of particular things, ac- this demand is not specific to the librarian but de- tualities, and specific kinds, we find that some are scribes, for example, the biologist and sexologist Al- alike, in general characters and in specific characteris- fred Kinsey equally well.) tics; and we may consequently relate them in a class, or classes, that is classify them. Shiyali Ramamrita Ranganathan (1967, CP:77-78 emphasis original) Jason E. L. Farradane (1950, 83). Ranganathan distinguished five senses of “classification.” Classification is a theory of the structure of knowl- 1. Division. (See Chap CC). edge, i.e. of the relations between different parts of This is the primitive meaning of the term “classificati- knowledge. No arbitrary method of grouping, how- on.” Even a child practises classification in Sense 1 ever carefully applied, is true classification. The prob- with its playthings. Even early man had practised it. lem is primarily epistemological. What is true knowl- 2. Assortment [grouping of things of the same sort]. edge, and what are true relations between the parts of (See Chap CD). knowledge? It is essential to define these if the classi- Classification in Sense 2 is inherent in Man. Perhaps it fication is to be true and logically sound. is a concomitant of the finiteness of the speed of neural impulses in the human body. When the speed is finite, structure emerges. Wherever there is a structure, Knowl. Org. 44(2017)No.2 125 Reviews of Concepts in Knowledge Organization

sequences emerge. When sequence is helpful to the Classification in Sense 2 has only a Scheme for Classes purpose at hand, it is Classification. The sequence ine- associated with it. But Classification in Sense 5 has a vitable inside of the skin, so to speak, gets expressed Scheme for Classification associated with it. We shall extraneurally also. To classify in Sense 2 is thus a neu- restrict the meaning of the term “Classification” to ral necessity. Sharpness in thinking, clarity in expressi- Classification in Sense 5. on, unerringness in communication, expedition in response, and exactness in service depend ultimately on Bonifatii Mikhailovich Kedrow (1975, 1:3 emphasis origi- helpful sequence or Classification in Sense 2. The nal) work of philosophers and of taxonomists in the field Der Klassifizierung der Wissenschaften bedeutet den of classification is generally restricted to Classification Zusammenhang der Wissenschaften, der in ihrer Stellung in in Sense 2. (See Chap CL). einer bestimmten Reihenfolge oder in einem System 3. Classification in sense 2 plus Representing each enti- entsprechend einigen allgemeinen Grundsätzen zum ty by an ordinal number taken out of a system of or- Ausdruck kommt. dinal numbers, designed to mechanise the maintenance of the sequence, This can perhaps be generalized in this way: Classifica- 1. Either when an entity has to be replaced after tion of objects means the display of connections be- having been taken out of its position; tween the objects in a certain order or in a system re- 2. Or when a new entity has to be interpolated or flecting certain basic principles. Kedrow found that extrapolated in the correct place in the sequence. the principle of historicism must govern all natural This ordinal number is the Class Number. (See Chap classifications (see section 4.2cγ Genetic/historicist CG and CM). approaches to classification). Classification in Sense 3 is usually practised by large business concerns having to handle a large number of Georg Klaus (1976, 628-629) commodities. The Customs Authorities too use Classi- Klassifikation. Verfahren zur Unterteilung eine Klasse fication in Sense 3 in their published list of commodi- K von Dingen usw. in Teilklassen” (628). “Die dialek- ties liable to customs duty. tisch-materialistische Einstellung zur Klassifikation be- 4. Classification in Sense 3 when complete assortment is steht also nicht etwa darin, dass sie im Gegensatz zur made of an amplified universe—that is, when the enti- antidialektischen die Berechtigung und den Wert von ties and the pseudo-entities arising in the process of Klassifikationen bestritte, sondern darin, dass sie die successive assortment stand arranged in one filiatory These von der zeitlichen und strukturellen Relativität sequence, each with its Class Number. (See Chap CH der Klassifikationen vertritt, während die antidialek- and CK). tische Auffassung die Klassifikationen in jeder Hin- Classification in Sense 4 is not used very much. It is sicht als absolut betrachtet. only classification in Sense 3 and Sense 5 that are fre- quently in demand. Phillip R. Sloan (1981, 68) 5. Classification in Sense 4 with all the entities re- The arrangement of objects or entities into groups or moved but only the pseudo-entities or classes re- classes, usually on the basis of perceived similarity and tained—each class having the number representing it. difference. (See Chap CM). It is classification in Sense 5 that is used, ISO 5127-6 (1988, 93 from Simões et al. 2016, 531; note 1. Either when the universe classified is infinite; that this standard has been revised by ISO 5127:2001) 2. Or when some of the entities are unknown and A classification system is an “indexing language in- unknowable at any moment, even though the uni- tended for a structured representation of documents verse classified is finite. or data, through the use of indexes and corresponding In particular, it is Classification in Sense 5 that is prac- terms, in order to allow systematic access, resorting to tised by the library profession. an alphabetical index, if necessary. It should be recalled that that in classification in Sense 5 1. The individual entities do not figure in the com- Lois Mai Chan (1994, 259) plete assortment; Classification is: 2. Classes take the place of entities ; and thus, The multistage process of deciding on a property or 3. Each class including the Original Universe is a characteristic of interest, distinguishing things or ob- Class of Classes. jects that possesses that property from those which lack it, and grouping things or objects that have the 126 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

property or characteristic in common into a class. ing, sorting, grouping, arranging, ordering, ranking, Other essential aspects of classification are establish- mapping and correlating. ing relationships among classes and making distinctions within classes to arrive at subclasses and finer di- Elaine Svenonius (2000, 10) visions. Organization can take many forms. Its prototypical form is classification. Classification brings like things Francis L. Miksa (1994, 144) together. In traditional classifications, like things are [Bibliographic] Classification is the activity of creating brought together with respect to one or more speci- categories into which bibliographic items of all kinds fied attributes. Any number of attributes can be used may be placed (i.e., the work of the classificationist) to form classes of documents embodying informa- and also the activity of identifying bibliographical items tion, such as same size or color, same subject, or same in terms of the categories already extant in a given sys- author. However, the most important attribute for a tem (i.e., the work of the classifier). It encompasses system whose objectives is to organize information is systems for arranging items on the shelves of libraries the attribute of “embodying the same work.” (sometimes called “bibliothecal” classification), as well as systems for arranging the surrogates of items in Hubert Feger (2001, 1966) catalogs (sometimes called “bibliographical” classifica- Classification is the assignment of objects to classes. tion). It includes classificatory systems based on all Later on, this was expanded (Feger 2015, 805): kinds of item characteristics (subject, form, author, ci- The fundamental goal of classification is to find struc- tation, size, etc.), in all forms of order (logical and sys- tures common to a group of objects, using properties tematic, alphabetical, faceted, etc.), with all kinds of to classify the objects into subgroups based on the operating methods (pre- and post-coordinated, statis- similarity of their properties. tically based clustering and identification, etc.), and differing in scope from the universal to the very narrow. The Portuguese Language Dictionary of the Academy of Sciences Finally, library classification embraces a wide range of (Academia das Ciências de Lisboa 2001, 837 cited in purposes, although most often its chief purpose has Simões et al. 2016, 531). been to facilitate document retrieval. Classification is the “action of distributing in classes, by categories ... according to precise criteria.” Frederick Wilfrid Lancaster (1998, 17) Classification is “sorting items into “conceptual classes” Ernst Mayr and Walter Joseph Bock (2002) and “forming classes of objects on the basis of their The logical consequence of the definition of class is subject matter.” that classification must be defined as the ordering of diversity into classes of similar entities. And this has Geoffrey Bowker and Susan Leigh Star (1999, 10) been traditionally the almost universally accepted con- A classification is a spatial, temporal, or spatio- cept of classification .… A classification of organisms temporal segmentation of the world. A “classification is based on the shared possession of their diverse at- system” is a set of boxes (metaphorical or literal) into tributes. The units of similarity in a Darwinian classifi- which things can be placed to perform some kind of cation are called taxonomic characters that have the work—bureaucratic or knowledge production. property of being homologous to one another in the several entities or groups. The claim of a few modern Arlene G. Taylor (1999, 237) authors that there is no agreement on the definition of The placing of subjects into categories; in organizing the word “classification” is quite misleading. Actually, of information, classification is the process of deter- prior to 1950 there was virtually total unanimity on the mining where an information package fits into a given usage (in classification) of the words classification and hierarchy and then assigning the notation associated class, as referring to the grouping of similar items. with the appropriate level of the hierarchy to the information package and its surrogate. A classification is defined as “The arrangement of entities in a hierarchical series of nested classes, in which M.P. Satija (2000, 222) similar or related classes at one hierarchical level are Classification means to divide objects/entities (both combined comprehensively into more inclusive classes abstract and concrete) on the basis of their differences at the next higher level” (176). or, conversely, the grouping of entities on the basis of their similarities. Classification is any process of divid- Knowl. Org. 44(2017)No.2 127 Reviews of Concepts in Knowledge Organization

Class—(in classification) A group or collection of en- facets. The classification of living things is a taxon- tities (individuals), possessing attributes or traits in omy. (The term taxonomy is increasingly used for any common (“being similar”), a kind or sort, grouped to- type of classification.) A classification is now often gether under a general or class name. called an ontology, particularly if it gives richer con- Classification—The arrangement of similar entities cept relationships. (objects) in a hierarchical series of nested classes, in which each more inclusive higher-level class is subdi- Faria and Pericão (2008, 258 in Simões et al. 2016, 531) vided comprehensively into less inclusive classes at the Classification is a “group of ordered concepts, distrib- next lower level. uted systematically in classes, forming a structure” and a Darwinian classification—The ordered grouping of “structuring of concepts into classes and subdivisions organisms into classes, according to their similarities to express the existing semantic relationships between and consistent with their inferred evolutionary history. them. Downward classification—Establishing groups by logical division. Clare Beghtol (2010, 1045) Evolutionary classification—A classification that duly To classify means to put things into meaningful groups. considers both evolutionary processes, the ecological Things can be physical objects, ideas, events, or any- adaptiveness of evolutionary divergence (degree of thing else that human beings can perceive or imagine, difference) and the genealogy (phylogeny) of the taxa. and a meaningful group can be formed using any char- Basically equal to a Darwinian classification. acteristic or combination of characteristics of the Hierarchical classification—The system of ranks that things. Groups can be considered to be permanent or indicates the categorical level (level of difference) of they can be considered temporary responses to a need each taxon (191). of the moment.

Kjeld Schmidt and Ina Wagner (2004, 392) Ingetraut Dahlberg (2010, 2941 list typography added) Classification, in turn, is a special practice of categoriza- With this journal [International Classification, 1974–1992, tion, involving pre-established and systematic systems thereafter Knowledge Organization], “classification” was of signs. That is, classification is a linguistic operation understood as a multi-meaning word that includes the of applying a classification scheme, i.e., an ordered set following concepts: of signs that is pre-established according to (a) some 1. classification in the sense of “classification system,” general principles and criteria of ordering and (b) some i.e., a system of classes arranged in hierarchical or fac- procedures of identification and naming. In short, an eted order; act of classification is an application of a classification 2. classification in the sense of classifying, i.e., estab- scheme. Classification systems (such as thesauri) can lishing a system of classes; thus be seen as instantiations of classification schemes. 3. classification in the sense of classing, i.e., relating the classes of a classification system to objects or sub- Classifications and categorizations are both convention- jects of reality; and based practices and equally so. But classifications are 4. classification in the sense of classification science, convention-based in a quite specific sense. In the case i.e., relating to this field of study and its activities. of categorization there are no pre-established principles and criteria for determining the correctness of an act of Oxford English Dictionary (update from 2010) categorization. With acts of classification, however, classification, n. such pre-established principles and criteria exist, in that Origin: A borrowing from Latin. Etymon: Latin classi- they specify relationships between items in terms of, for ficatio. example, class/ membership, part/whole, composition, Etymology: < post-classical Latin classificatio (1673 in cause/effect, origin/fate, function, ownership, value/ a German source; 1767 in Linnaeus) > classical Latin risk, location, or state. Accordingly, an actor applying a classis class n. + -ficātiōn- , -ficātiō -fication suffix. classification scheme in a particular case can be held ac- Compare German Klassifikation (1760 or earlier as countable in terms of the principles, criteria, and proce- †Classification), Swedish klassifikation (1740 as †clas- dures of the classification scheme. sification), Danish klassifikation (1748), French classification (1780), Italian classificazione (1796). Compare Dagobert Soergel (2004, 358) slightly later classify v. A classification is a structure that organizes concepts 1. The result of classifying; a systematic distribution, into a meaningful hierarchy, possibly in a scheme of allocation, or arrangement of things in a number of 128 Knowl. Org. 44(2017)No.2 Reviews of Concepts in Knowledge Organization

distinct classes, according to shared characteristics or ous than them.10 This is also the result of this opera- perceived or deduced affinities. Also: a system or tion. We want, as much as it is possible, this result to method for classifying. be constant, i.e. the classification must remain stable 2. The action of classifying or arranging in classes, ac- for a little transformation11 of data. cording to shared characteristics or perceived affinities; Note 10: assignment to an appropriate class or classes. In the case of infinite classifications, this requirement, 3. A category to which something is assigned; a class.” of course, must be weakened: we may only want the (infinite) cardinal of the classification to be less than Adriano Veloso and Wagner Meira (2011, 9) or equal to the (infinite) cardinal of the set of objects In a classification problem, there is a set of input- to be classified. output pairs (also referred to as instances or examples) Note 11: of the form zi = (xi; yi): Each input xi is a fixed-length The sense of it will have to become clearer. record of the form ‹a1, . . . ,al›; where each ai is an attribute-value. Each output yi draws its value from a discrete WordNet (3.1) [downloaded 2016-05-21] defines four and finite set of possibilities y = {c1, . . .,cp}, and indi- senses of the noun “classification” of which three are cates the class to which zi belongs. Cases where yi = ? relevant for this entry: indicate that the correct class of zi is unknown. There is “– S: (n) categorization, categorisation, classifica- a fixed but unknown conditional probability distribution tion, compartmentalization, compartmentalisation, as- P(y|x), that is, the relationship between inputs and out- sortment (the act of distributing things into classes or puts is fixed but unknown” categories of the same type) (11): This formulation implies that the classification – S: (n) classification, categorization, categorisa- problem corresponds to the problem of function ap- tion (a group of people or things arranged by class or proximation. category) – S: (n) classification, categorization, categorisa- Daniel Parrochia and Pierre Neuville (2013, 21) tion, sorting (the basic cognitive process of arranging Definition 1.9.1 We call “classification” the operation into classes or categories).” consisting of sharing, distributing or allocating objects in classes or groups which are, in general, less numer-

Knowl. Org. 44(2017)No.2 129 Gems from our Digitization Project

Gems from our Digitization Project

Reprinted from International Classification: A Journal Devoted to Concept Theory, Organization of Knowledge and Data, and to Systematic Terminology, vol. 7 (1980) No. 1. The masthead identifies the “editors” as: Ingetraut Dahlberg, Alwin Diemer, Arashanipalal Neelameghan and Jean M. Per- reault. The article is reprinted without editorial interpola- tion—Ed.-in-chief.

Douglas J. Foskett University of London Library, England

Systems Theory and its Relevance to Documentary Classification

Abstract: In view of the impact of systems theory for the construction of classification systems the two major contributions of Dewey are summarized as well as the new methods of facet analysis and organization brought into classification by Ranganathan. With the latter’s “canonical” solution for the contents and arrangement of main classes, however, contemporary philosophical thought regarding the organization of knowledge seems to have been neglected. The work of the Classification Research Group and elsewhere considering integrative level theory will improve the science of classification systems construction. Besides this the influence from psychology and The two major contributions of Melvil Dewey are in this linguistics on the recognition of relationships between same tradition. In 1870, the current dominant philosophy concepts is outlined as well as some practical implications was the result of combining Aristotelian logic with em- of the systems approach on classification. (I.C.) pirical investigation of nature in the classificatory sciences; this gave Dewey the idea of hierarchical subdivi- Foskett, D. J.: Systems theory and its relevance to docu- sion of subjects and their relative location on library mentary classification. In: Intern. Classificat. 7 (1980) No. shelves, replacing the fixed location of specific books. 1, p. 2-5. From mathematics he took the decimal fraction notation, which admirably reflects hierarchical subdivision and the 1.0 Dewey’s approach subordination of subjects:

The history of classifications of knowledge shows that 599 Mammals schemes for the ordering of knowledge or of documents 599.8 Primates containing knowledge always, and inevitably, reflect the 599.88 Apes philosophies and theories of knowledge which are domi- 599.884 Gorillas nant at the time. H. E. Bliss called it the “educational and scientific consensus.” It need not surprise us: if a phi- Dewey also realised that hierarchical subdivision was not losophy has a social function, and I believe it has, it is sufficient by itself, and introduced what he rightly called a precisely to provide a method for investigating the struc- “mnemonic principle” for subdividing geographically by ture of knowledge in order to understand the world the use of numbers taken from class 900, and also by his about us. And once we start speaking of “structure,” we “form divisions” for dictionaries, periodicals and so on. are in the realm of classification. Even in his first edition, he noted that “users of the 130 Knowl. Org. 44(2017)No.2 Gems from our Digitization Project scheme will notice this mnemonic principle in several variant and synonym, and as many bound term com- hundred places in the classification.” pounds as the compiler may chance to come across in the Dewey calls his principle “mnemonic”; we now call it literature of his subject, no matter what the cost. Some “synthesis,” and it has been developed to a high degree in attention to relations and categories has been forced on the UDC and by Bliss in his Bibliographic Classification. the compilers, as is shown by the ceaseless activity of al- Ranganathan was the first to develop a true theory of tering, and publishing so-called “revised editions.” It is analytico-synthetic classification, and his system of facet unfortunate that this “might is right” philosophy has had analysis went so far to meet the needs of ordering and so much influence through the sheer weight of American indexing the complex subjects of modern documentation publications. that it has passed into the common stock of professional In Europe, we have a much longer tradition of intel- knowledge, and many people who now speak confidently lectual analysis, and some spirited resistance has occurred. of facet analysis have never heard of Ranganathan. As To give but one example: the EUDISED Thesaurus Goethe said, “Die Tat ist alles, nicht der Ruhm.” compiled by Jean Viet for the Council of Europe has a faceted structure which is immediately understood by us- 2.0 Ranganathan’s method ers in many different countries; the ERIC Thesaurus of the United States Office of Education, despite its 6 edi- Like all epoch-making discoveries, Ranganathan’s method tions, continues to earn harsh damnation even from insti- was simple: he showed that a classification scheme could tutions within the ERIC system. incorporate hierarchical subdivision of classes—a most valuable aid to research, as Sandison has recently con- 3.0 The problem of the “Main Class” firmed (1) —into a framework which kept in separate schedules those terms which related in different ways to But facet analysis does not offer a solution to another their Main Class. In his Colon Classification, these are major problem of classification: the choice of Main terms which represent categories of Matter and Energy, Classes. Facet analysis requires a starting point, a named and they are separate from each other and from terms and defined area of knowledge, a Main Class in which the which represent Space, or geographical division, and technique can be applied. Ranganathan avoided attacking Time, or chronological division. This method released this problem on the grounds that there are recognised schemes of classification from the straitjacket of “bound “canonical,” Main Classes, and he had other more urgent terms,” that is, hierarchies in which subdivisions of a questions to answer. But he acknowledged the need for class derived by different characteristics are listed in the something more than tradition by his introduction of same schedule, as if they were derived by the same char- what he called “Basic Classes.” These are in effect any acteristic. For example, consider this array from the subjects that a compiler may choose to name as starting UDC: points, the type of special subject for which the British Classification Research Group has been making faceted 37 Education classifications for more than a quarter of a century. This 37.018 Fundamental forms of education is satisfactory as far as it goes, and indeed has been a fer- 37.018.2 School education tile source of ideas on concept analysis, relational analy- 37.018.26 Attitudes of parents to school sis, and several problems connected with the choice and 37.018.263 Parent-teacher relations ordering of terms within facets. But fundamentally it is a pragmatic approach, and so more or less subjective. Cer- It is obvious that, unlike the single hierarchical array from tainly, we cannot escape the subjective in a matter like the Dewey above, this supposedly single hierarchy in fact structure of knowledge, but I believe that we have so far presents a mixture of several characteristics: schools, par- made little progress in resolving the main issue precisely ents, attitudes, are all terms which belong to different ar- because we rarely attempt to reflect current dominant eas of knowledge. They are not a hierarchy, but are philosophies. We do not take enough notice of what con- bound together as if they were. temporary philosophers and scientists have to say about Facet analysis thus provides a complete solution to the nature of knowledge. one of two major problems in documentary classifica- A few centres have been remedying the situation. The tion. No modern scheme is without it, and we can also FID/CR Committee, because of its close connection find recognition in thesaurus construction, even where it with the UDC, has given some attention to the problems is ignored or disguised, as in most American thesauri, of general classifications. The British CRG has provided which attempt to solve indexing problems by the steam- the factory of ideas for the PRECIS system of indexing hammer method of including every conceivable term, used in the British National Bibliography and for the new Knowl. Org. 44(2017)No.2 131 Gems from our Digitization Project edition of Bliss’s Bibliographic Classification under the to natural entities; central to this is the concept of a direction of Jack Mills. The Seminars of the DRTC in “whole,” something which has a discernible identity and Bangalore continue and enlarge the work of Rangana- can be distinguished in isolation. This line of thought al- than, and the three International Study Conferences on so offered an explanation of Ranganathan’s concept of Classification Research (Dorking 195 7, Elsinore 1964, “‘personality” which was more detailed than any which he Bombay 1975) have been notable landmarks; the Third in himself gave. It thus fitted in very neatly with the tech- particular contains several papers relevant to my present nique of facet analysis. theme (2). This is particularly significant because that In his paper to the Bombay Conference, Eric de Conference took the perspective of “global information Grolier does less than justice to these ideas (8). He dis- networks,” which of necessity involves considering the misses Derek Austin’s NATO Project, which was wholly whole universe of knowledge and not special subject ar- a CRG project, but pays tribute to the work of J. L. Jolley eas in isolation from one another. In my book on the so- and A. J. Mayne, both of whom were CRG members, and cial sciences·(3) I drew attention to the difficulty, in mak- certainly Jolley’s concept of the “holotheme” relates ing a special subject scheme, of knowing how and where closely to the theory of integrative levels. to stop drawing on terms from marginal fields. It is true that, in its original formulation twenty years This problem is entirely a matter of the relationships, ago (9), the theory concentrated on “things,” because this in real life, between concepts. These may be of two main seemed the simplest way to relate it to Ranganathan’s types: for convenience, I shall call them basic or primary, concept of a Personality facet: “the basis, the host, the and occasional or secondary. The basic relations, which locus of all other fundamental categories.” But of course correspond more or less to what J.-C. Gardin calls “para- we never assumed that things existed in total isolation digmatic relations” are those which maintain the identity from all other natural phenomena. Taken in turn as a se- of a concept and are part of what J. E. Farradane calls its ries of Personality facets, Things attract to themselves a “unique definition.” The occasional relations are those similar series of Matter and Energy facets. The theory which come into being as part of a particular set of phe- thus readily meets de Grolier’s criticism that is does not nomena which are not necessary to the existence of the deal with the ordering of social fields or activities. What concept, but may affect it. A human being is always a ver- it does is to relate these activities to the very entities tebrate mammal; a human being may have red hair, or which engage in them; one can certainly have the concept engage in professional conferences, but neither of these of an activity, just as Ranganathan has the concept of an are essential attributes without which the being could not Energy facet, but in-the real world activities are no more exist as human. and no less than the mode of existence of things, and in- We are therefore inextricably involved with the process deed things and their activities are inseparable. of concept formation, and I have put forward some pre- This has been demonstrated by the now large body of liminary thoughts on this, some years ago, in a paper on material, published mainly in the USA, of which Berta- “User psychology” (4). Some very important recent work lanffy, Kenneth Boulding and Ervin Laszlo are among has been published by Ingetraut Dahlberg, first in her the chief contributors. The idea of a “system” is any en- contribution to the Bombay Conference, and more fully tity whose characteristics are identified as the nature of in her Ranganathan Lectures in Bangalore (5); the latter, its parts and the relations between them. A bicycle is perhaps for the first time since H. E. Bliss, discuss in de- more than a heap of bits of metal, rubber, plastic, and so tail the question of the organisation of knowledge on; the relationships set up between these parts trans- through the medium of general schemes of classification. forms the heap into the characteristic appearance of a bicycle and enables it to perform the characteristic function 4.0 The Contribution of “General System Theory” of a bicycle by converting the rotary motion of the ped- als into the horizontal motion of bicycle and passenger My paper here is an attempt to add to this line of along the road. A Committee is more than a collection of thought by discussing some ideas derived from General single individuals: they group themselves in a specified re- System Theory. There are many works on this, but the lationship, elect a chairman, address their comments to basic text, in my view, is that of Bertalanffy, General System the chairman, and take collective decisions binding on all Theory (6). However, the basic ideas were first discussed of them. In fact, the activities of any system are just as by the CRG in the late 1950s, through a paper by Joseph essential a feature as its constituents. Needham dating back to 1937, his Herbert Spencer lec- A system may also be a constituent part of another tures given to the University of Oxford (7). The idea of system of a higher order of organisation. Thus a word is “integrative levels” in nature seemed to provide a clue to a system of letters organised in a certain way—their se- an objective method of ordering concepts which related quence. A sentence is a System of words organized in a 132 Knowl. Org. 44(2017)No.2 Gems from our Digitization Project particular sequence, and a paragraph is a system of sen- technological and social and they react on, and are re- tences. A book is a system of paragraphs and a library is acted on by, his individual self. On this view, Laszlo has a system of books. A classification conference is a system no difficulty in refuting the common objection of deter- of classificationists. Thus we have, in the real world minism, and showed conclusively that systems philosophy which provides the subjects for documentation, a system encompasses social and human value in a “framework for of systems in an order of increasing complexity of parts a nonnative ethics.” In a letter to me, he agreed that my and relations. Applying this concept to schemes of classi- Sayers volume paper was completely in accordance with fication will produce an ordered system which strongly his own ideas, and indeed extended them into a new resembles the scheme produced in outline by Ingetraut area—documentary classification. Dahlberg in her Ranganathan Lectures. The notion of a series of systems integrated by in- 5.0 The influence from psychology and [l]inguistics creasing complexity of organisation is not new in the natural sciences; it is implicit in the work of Auguste Recent work in two other major fields, which serve to il- Comte, to go back no more than 150 years. The series of lustrate the interpenetration of science and the humani- Fundamental particles—atoms—molecules—masses, is ties, follows a similar path: psychology and linguistics. In universally accepted. Whether the notion can be carried psychology, I. Dahlberg has drawn attention to the essen- throughout the whole field of knowledge remains in dis- tial basis of concept formation, with reference to Ger- pute. In the CRG, for example, D. W. Langridge has con- man literature, and I have drawn on the work of leading sistently claimed that one cannot apply the idea of levels psychologists, notably J. P. Guilford and Jean Piaget, in to the Humanities (10), and performs an extremely useful my literature review on “Informatics” (12). Guilford’s service relevant to this paper. in analysing the theories of “structnre of intellect” model also influenced J. E. Far- several contemporary philosophers concerning the struc- radane’s well-known work in relational analysis. Piaget has ture of knowledge. His principal objection is that systems shown, through a long series of books, that concept theory implies that natural science is the paradigm of all formation proceeds by the assimilation of data given by knowledge and that what holds good for ordering knowl- the senses, through observation and experiment, into a edge in the sciences must apply to all the other areas of structure of concepts already formed in the mind of the thought. learner; through the study of growing children, he and This objection certainly applies to a mechanical trans- his co-workers proved that it is by this process of classifi- fer of particular theories in science to the other areas, but cation that infants begin to develop the ability to cope that is not my view of systems theory. A “general” theory with their environment. Teachers all over the world have can only be general if it can indeed apply through all learned how to teach through study of these works. Pia- fields; this is what makes it general, and generalisations get has also contributed to the philosophy of Structural- have been the main aim of philosophers and scientists ism. “In short,” he writes, “the notion of structure is throughout the ages. The crucial test of any theory is the comprised of three key ideas: the idea of wholeness, the extent of its application, and a theory is replaced when idea of transformation, and the idea of self-regulation” another theory is proved to account for a wider range of (13). He applies the notion to the whole of knowledge, phenomena. and it is not difficult to see that it has direct resemblances Langridge is right, however, when he claims that more to systems theory and with documentary classification. investigation is needed. Much of the ground has been In order to achieve communication, concepts in the covered by Ervin Laszlo, who does extend systems the- mind of an author have to be expressed in a form which a ory to the Humanities (11). His aim was, not to refute the reader or listener can understand, and everywhere there are theories of other philosophers, but to collate or map barriers. Piaget himself relates his work to linguistics and them into “a common, internally consistent framework received some critical comments in what I regard as a wherein their particular propositions become mutually re- seminal work in this field, Thought and Language by L. S. Vy- inforcing as descriptions and explanations of one reality gotsky (14). First published in Moscow in 1934, it had with a rationally knowable, overarching species of order.” hardly any impact until an English translation was pro- By considering Man himself as a cognitive system, we can duced by the Massachusetts Institute of Technology in see that he exists as an individual by virtue of two sets of 1962, with an introduction by Jerome S. Bruner. In Vygot- relationships: those internal to his own individual body, sky’s view, the crucial activity in concept formation is the which become progressively organised through his own transforming of “spontaneous concepts” into “scientific personal experience, and those external to him, which concepts” by incorporating sense-data derived from the consists of the world or environment in which he finds environment into a network of related concepts already in himself. These external relations are physical, biological, the mind, and expressing them in “units of verbal Knowl. Org. 44(2017)No.2 133 Gems from our Digitization Project thought,” or “word-meanings,” which combine scientific the way in which even the same specialist approaches the thoughts with units of speech and so become communica- literature may-vary from one occasion to another. But the ble. Word combinations form sentences, and “just as the order must make sense: the specialist must be able to rec- sense of a word is connected with the whole word, and not ognize the basis for the order, hence the incentive to re- with its single sounds, the sense of a sentence is connected flect the current dominant philosophy. Specialists need with the whole sentence, and not with its individual and know about classification as an intellectual tool for words.” their work; witness the success of the Classification Soci- I give these two major examples to illustrate what I have ety and of the Gesellschaft für Klassifikation. These spe- described more fully in “Informatics,” namely, that we can cialists look at knowledge from the point of view of their find the basic concepts of systems theory in the works of own subject; only librarians and information officers look leading modern thinkers in a wide range of subjects. They at classification from the perspective of the whole uni- are also related, as both Piaget and Vygotsky acknowledge, verse of knowledge. A scheme for documentary classifi- to the philosophical aspects of dialectical materialism as cation must therefore be more than merely a collection developed by Marx and, more particularly by Engels in his of specialist schemes: this would not be a system in itself, Dialectics of Nature. There is plenty of evidence to show it would be no more than a heap of unrelated parts. that we can cover the whole knowledge by relating subject Systems theory says that internal relations between the analysis, or classification, to a general theory of systems. parts are essential if these parts are to have the organisa- In a general classification for documentation, any sys- tion of an entity capable of existence as an integral whole tem can be named a Basic Class, in DRTC terms, because in a particular environment. In our case, the environment all systems can be analysed by facet analysis. The system is the library and information service and the documents itself, considered as a whole, becomes the Personality. Its it contains; our aim in classifying is to reflect and demon- constituent parts and the relations between them become strate the order and harmony existing in the real world, the Matter and Energy, which I will call Energy A. The the universe of nature, including the world of Man. This relations of the system with its environment are also pro- is what writers write about from their own experience, cesses, which I will call Energy B. The other sys tems in and this forms the contents of the documents we have to the environment, which react with our original system, organise. The record of the thought is always incomplete, are Agents or, in Ranganathan’s own terms, Second always changing, always advancing. Round Personality. Of course, we do not have to accept The aim of scientists and philosophers is to find ex- Ranganathan’s terms; I do so here in order to illustrate planations, or “laws of nature” which can be used to our how appropriately systems theory fits the scheme of the advantage in our never-ending struggle to master our en- greatest contributor to documentary classification since vironment. Knowledge advances not only by more and Bliss and Dewey. The work of the CRG and of many more detailed analyses of individual subjects in isolation, compilers of thesauri demonstrate that the fit is even but by the formulation of more general principles and more obvious if the categories or terms used are chosen explanations with wider and wider application. Classifica- on a pragmatic basis to suit each subject field, without be- tion theory must take the same path. ing related to any set of fundamental categories. From the point of view of the foundations of general References: classification schemes, moreover, we gain little from criti- cisms of any scheme on a purely empirical basis, asking (1) Sandison, A.: The SRL Classification for books on only questions like “What has been omitted?” or “What shelves. In: J. of Librarianship, 12, (1980) 1, p. 26-41. has been placed in the wrong schedule?” This sort of (2) Neelameghan, A., (ed.): Ordering systems for global unproductive approach disfigures some of the articles on information networks. Bangalore, D.R.T.C., 1979. the UNISIST Broad System of Ordering in the recent is- (3) Foskett, D. J.: Classification and indexing in the so- sue of the FID journal, International Forum on Information cial sciences. London, Butterworths 2nd ed., l974. and Documentation (15). (4) Foskett, D. J.: User psychology. In: International Con- ference on training for information work, Rome, No- 6.0 Practical implications of the systems approach vember 1971, ed. by Georgette Lubbock. Italian Na- tional Information Institute and FID, 1972. What, then, are the practical implications of the systems (5) Dahlberg, I.: On the theory of the concept. In: Nee- approach to documentary classification? The main pur- lameghan, A., ed., vide supra (2). pose of any scheme of classification is to order docu- - Ontical structures and universal classification. Banga- ments in a way which makes sense to specialists in each lore: Sarada Ranganathan Endowment for Library field. It may not always be the most useful order, because Science, 1978. 134 Knowl. Org. 44(2017)No.2 Gems from our Digitization Project

(6) Bertalanffy, L. von: General System Theory: founda- (11) Laszlo, E.: Introduction to systems philosophy: to- tions, development, applications. New York: Bra- ward a new paradigm of contemporary thought. ziller 1968, and London: Penguin Press 1971. New York: Gordon and Breach 1972. (7) Needham, J.: Time, the refreshing river. London: Al- (12) Foskett, D. J.: Progress in documentation: ‘Informat- len and Unwin 1943. ics’. In: J. of Documentation, 26 (1970) No. 4, p. 340- (8) Grolier, E. de: In search of an objective basis for 369. the organization of knowledge. In: Neelameghan, (13) Piaget, J.: Structuralism. London: Routledge and A., ed., vide supra (2). Kegan Paul1971. (9) Foskett, D. J.: Classification and integrative levels. In: (14) Vygotsky, L. S.: Thought and Language. Cambridge The Sayers Memorial Volume. London: Library As- Mass.: M. I. T. Press, 1962. sociation 1961, p. 136-150. (15) International Forum on Information and Documentation, 4 (10) Langridge, D. W.: Classification and indexing in the (1979) No. 3. humanities. London: Butterworths 1976.

Knowl. Org. 44(2017)No.2 135 Books Recently Published

Books Recently Published

Henderson, Margaret E. 2017. Data Management: A Practi- Schmidt, Krista and Tim Carstens. 2017. The Subject Liai- cal Guide for Librarians. Lanham, Maryland: Rowman & son's Survival Guide to Technical Services. Chicago: ALA Littlefield. Editions. Library of Congress. 2017. The Card Catalog: Books, Cards, and Literary Treasures. San Francisco: Chronicle Books.

Knowl. Org. 44(2017)No.2

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

Publisher Reference citations within the text should have the form: (Author year). For example, (Jones 1990). Specific page numbers are required for quoted ERGON-Verlag GmbH, Keesburgstr. 11, D-97074 Würzburg material, e.g. (Jones 1990, 100). A citation with two authors would read Phone: +49 (0)931 280084; FAX +49 (0)931 282872 (Jones and Smith, 1990); three or more authors would be: (Jones et al., 1990). E-mail: [email protected]; http://www.ergon-verlag.de When the author is mentioned in the text, only the date and optional page number should appear in parentheses – e.g. “According to Jones (1990), …” Editor-in-chief (Editorial office) or “Smith wrote (2010, 146): ….” A subsequent page reference to the same cited work (e.g., to Smith 2010) should have the form “(229).” KNOWLEDGE ORGANIZATION In-text citations should not be routinely placed at the end of a sentence Journal of the International Society for Knowledge Organization or after a quotation, but an attempt should be made to work them into the Richard P. Smiraglia, Editor-in-Chief narrative. For example: [email protected] “Jones (2010, 114) reported statistically significant results. “Many authors report similar data; according to Matthews (2014, 94): “all Instructions for Authors seven studies report means within ±5%.” References should be listed alphabetically by author at the end of the ar- Manuscripts should be submitted electronically (in Word format) in English ticle. Reference lists should not contain references to works not cited in the only via ScholarOne at https://mc04.manuscriptcentral.com/jisko. Manu- text. Author names should be given as found in the sources (not abbrevi- scripts that do not adhere to these guidelines will be returned to the authors ated). Journal titles should not be abbreviated. Multiple citations to works by for resubmission in proper form. the same author should be listed chronologically and should each include the Manuscripts should be accompanied by an indicative abstract of ap- author’s name. Articles appearing in the same year should have the following proximately 250 words. Manuscripts of articles should fall within the range format: “Jones 2005a, Jones 2005b, etc.” Journal issue numbers are given 6,000-10,000 words. Longer manuscripts will be considered on consultation only when a journal volume is not through-paginated. References for pub- with the editor-in-chief. lished electronic resources should be accompanied by either a URL or DOI; A separate title page should include the article title and the author’s access dates are not required. Unpublished electronic resources may use an name, postal address, and E-mail address. Only the title of the article should access date in lieu of a data of publication. In cases of doubt, authors are appear on the first page of the text. Contact information must be present for encouraged to consult The Chicago Manual of Style 16th ed. (or online), author- all authors of a manuscript. date reference system (chapter 15). To protect anonymity, the author’s name should not appear on the Examples: manuscript. Dahlberg, Ingetraut. 1978. “A Referent-Oriented, Analytical Concept Criteria for acceptance will be appropriateness to the field of knowledge Theory for INTERCONCEPT.” International Classification 5: 142-51. organization (see Scope and Aims), taking into account the merit of the con- Howarth, Lynne C. 2003. “Designing a Common Namespace for Search- tents and presentation. It is expected that all successful manuscripts will be ing Metadata-Enabled Knowledge Repositories: An International Perspec- well-situated in the domain of knowledge organization, and will cite all rele- tive.” Cataloging & Classification Quarterly 37, nos. 1/2: 173-85. vant literature from within the domain. Authors are encouraged to use the Pogorelec, Andrej and Alenka Šauperl. 2006. “The Alternative Model of KO literature database at http://www.isko.org/lit.html. Classification of Belles-Lettres in Libraries.” Knowledge Organization 33: 204- The manuscript should be concise and should conform to professional 14. standards of English usage and grammar. Authors whose native language is Schallier, Wouter. 2004. “On the Razor’s Edge: Between Local and Over- not English are encouraged to make use of professional academic English- all Needs in Knowledge Organization.” In Knowledge Organization and the language proofreading services. We recommend Vulpine Academic Services Global Information Society: Proceedings of the Eighth International ISKO Conference ([email protected]). 13-16 July 2004 London, UK, edited by Ia C. McIlwaine. Advances in knowl- Manuscripts are received with the understanding that they have not been edge organization 9. Würzburg: Ergon Verlag, 269-74. previously published, are not being submitted for publication elsewhere, and Smiraglia, Richard P. 2001. The Nature of ‘a Work’: Implications for the Or- that if the work received official sponsorship, it has been duly released for ganization of Knowledge. Lanham, Md.: Scarecrow. publication. Submissions are refereed, and authors will usually be notified Smiraglia, Richard P. 2005. “Instantiation: Toward a Theory.” In Data, In- within 6 to 8 weeks. formation, and Knowledge in a Networked World; Annual Conference of the Canadian The entire manuscript should be double-spaced, including notes and ref- Association for Information Science … London, Ontario, June 2-4 2005, ed. Liwen erences. Vaughan. http://www.cais-acsi.ca/2005proceedings.htm. The text should be structured with numbered subheadings. It should Upon acceptance of a manuscript for publication, authors must provide contain an introduction, giving an overview and stating the purpose, a main a digital photo and a one-paragraph biographical sketch (fewer than 100 body, describing in sufficient detail the materials or methods used and the re- words). The photograph should be scanned with a minimum resolution of sults or systems developed, and a conclusion or summary. 600 dpi and saved as a .jpg file. Footnotes are not allowed. Endnotes are accepted only in rare cases and should be limited in number; all narration should be included in the text of the article. Paragraphs should include a topic sentence and developed narrative; a typical paragraph has several sentences. Advertising Italics are permitted only for phrases from languages other than English, and for the titles of published works. Responsible for advertising: ERGON-Verlag GmbH, Keesburgstr. 11, 97074 Bold type is not permitted. Würzburg (Germany). Em-dashes should not be used as substitutes for commas. © 2017 by ERGON-Verlag GmbH. Illustrations should be embedded within the document. Photographs All Rights reserved. (including color and half-tone) should be scanned with a minimum resolution of 600 dpi and saved as .jpg files. Tables should contain a number and KO is published by ERGON-Verlag GmbH. caption at the bottom, and all columns and rows should have headings. All – The price for the print version (8 issues/ann.) is € 329,00/ann. in- illustrations should be cited in the text as Figure 1, Figure 2, etc. or Table 1, cluding airmail delivery. Table 2, etc. – The price for the print version plus access to the online version Author-generated keywords are not permitted. (PDF) is € 359,00/ann. including airmail delivery. Knowl. Org. 44(2017)No.2

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

Scope Aims

The more scientific data is generated in the impetuous present times, Thus, KNOWLEDGE ORGANIZATION is a forum for all those in- the more ordering energy needs to be expended to control these data in terested in the organization of knowledge on a universal or a domain- a retrievable fashion. With the abundance of knowledge now available specific scale, using concept-analytical or concept-synthetical approaches, as well as quantitative and qualitative methodologies. the questions of new solutions to the ordering problem and thus of im- KNOWLEDGE ORGANIZATION also addresses the intellectual proved classification systems, methods and procedures have acquired and automatic compilation and use of classification systems and thesauri unforeseen significance. For many years now they have been the focus in all fields of knowledge, with special attention being given to the prob- of interest of information scientists the world over. lems of terminology. Until recently, the special literature relevant to classification was KNOWLEDGE ORGANIZATION publishes original articles, published in piecemeal fashion, scattered over the numerous technical reports on conferences and similar communications, as well as book reviews, letters to the editor, and an extensive annotated bibliography of journals serving the experts of the various fields such as: recent classification and indexing literature.

KNOWLEDGE ORGANIZATION should therefore be available philosophy and science of science at every university and research library of every country, at every infor- science policy and science organization mation center, at colleges and schools of library and information sci- mathematics, statistics and computer science ence, in the hands of everybody interested in the fields mentioned library and information science above and thus also at every office for updating information on any archivistics and museology topic related to the problems of order in our information-flooded times. journalism and communication science KNOWLEDGE ORGANIZATION was founded in 1973 by an industrial products and commodity science international group of scholars with a consulting board of editors representing the world’s regions, the special classification fields, and the sub- terminology, lexicography and linguistics ject areas involved. From 1974-1980 it was published by K.G. Saur Ver- lag, München. Back issues of 1978-1992 are available from ERGON- Beginning in 1974, KNOWLEDGE ORGANIZATION (formerly Verlag, too. INTERNATIONAL CLASSIFICATION) has been serving as a As of 1989, KNOWLEDGE ORGANIZATION has become the common platform for the discussion of both theoretical background official organ of the INTERNATIONAL SOCIETY FOR KNOWL- questions and practical application problems in many areas of concern. EDGE ORGANIZATION (ISKO) and is included for every ISKO- member, personal or institutional in the membership fee. In each issue experts from many countries comment on questions of an adequate structuring and construction of ordering systems and on the Rates: From 2015 on for 8 issues/ann. (including indexes) € 329,00 (forwarding costs included) for the print version resp. € 359,00 problems of their use in opening the information contents of new litera- for the print version plus access to the online version (PDF). Member- ture, of data collections and survey, of tabular works and of other ob- ship rates see above. jects of scientific interest. Their contributions have been concerned with ERGON-Verlag GmbH, Keesburgstr. 11, D-97074 Würzburg; Phone: +49 (0)931 280084; FAX +49 (0)931 282872; E-mail: ser- (1) clarifying the theoretical foundations (general ordering theory/ [email protected]; http://www.ergon-verlag.de science, theoretical bases of classification, data analysis and re- Founded under the title International Classification in 1974 by Dr. duction) Ingetraut Dahlberg, the founding president of ISKO. Dr. Dahlberg (2) describing practical operations connected with indexing/classifi- served as the journal’s editor from 1974 to 1997, and as its publisher (Indeks Verlag of Frankfurt) from 1981 to 1997. cation, as well as applications of classification systems and thesauri, manual and machine indexing The contents of the journal are indexed and abstracted in Social Sci- ences Citation Index, Web of Science, Information Science Abstracts, INSPEC, (3) tracing the history of classification knowledge and methodology Library and Information Science Abstracts (LISA), Library, Information Science (4) discussing questions of education and training in classification & Technology Abstracts (EBSCO), Library Literature and Information Science (5) concerning themselves with the problems of terminology in gen- (Wilson), PASCAL, Referativnyi Zhurnal Informatika, and Sociological Ab- eral and with respect to special fields. stracts.