DOCSLIB.ORG

Ontoclean 2.0: a Framework for Rigidity

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Ontoclean 2.0: a Framework for Rigidity RC23754 (W0510-141) October 21, 2005 Computer Science IBM Research Report Towards OntoClean 2.0: A Framework for Rigidity Christopher Welty IBM Research Division Thomas J. Watson Research Center P.O. Box 704 Yorktown Heights, NY 10598 William Andersen Ontology Works, Inc. Research Division Almaden - Austin - Beijing - Haifa - India - T. J. Watson - Tokyo - Zurich LIMITED DISTRIBUTION NOTICE: This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. I thas been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g ,. payment of royalties). Copies may be requested from IBM T. J. Watson Research Center , P. O. Box 218, Yorktown Heights, NY 10598 USA (email: [email protected]). Some reports are available on the internet at http://domino.watson.ibm.com/library/CyberDig.nsf/home . Towards OntoClean 2.0: A framework for rigidity Christopher Welty IBM Watson Research Center 19 Skyline Dr., Hawthorne, NY 10532 [email protected] William Andersen Ontology Works, Inc. [email protected] Abstract A common criticism of OntoClean is that it relies on quanti- fied modal logic (S5) for its expression. While modal logic The OntoClean methodology was based on a set of is required for the expression of the rigidity axioms we pre- formal meta-properties whose semantics were sent here, many useful theorems derived from them may be specified in S5 modal logic. One of these used in non-modal settings. For this reason we give the be- metaproperties, Rigidity, has come under more fo- ginnings of a formal ontology of rigidity metaproperties cused scrutiny by the ontology community, and suitable for application with non-modal ontology authoring several problems with the formalization have been languages that permit quantification over relations. The discussed along with several solutions. In this pa- main point of this paper is to consolidate subsequent work per, we attempt to reconcile these results in a larger by many authors on the OntoClean meta-property called framework that exposes different kinds of rigidity, Rigidity, and establish a formal framework that exposes as well as two new metaproperties, actuality and differences between them. This work can be seen as a for- permanence, which deal more specifically with the mal ontology of rigid properties. behavior of properties with respect to time and ex- istence. 2 Background 1 Introduction One of the most often heard criticisms of the OntoClean methodology is that it requires modal logic. The truth is The notion of rigidity in the OntoClean methodology was that one does not need modal logic, nor modal logic reason- originally introduced to account for the conditions under ing, to use OntoClean in ontology-based systems. The for- which exemplification of properties by their instances is malizations of the OntoClean meta-properties were present necessary or essential1. Since then, many authors have fo- to clearly communicate their semantics, much like specify- cused on the analysis of rigidity in various settings, accom- ing a model theory for a language, and not with the intention panied by claims that the original definition failed to capture of being used in reasoning systems themselves. The formal- key elements such as time and actual existence. ization helped to maintain a level of rigor that can, in gen- The purpose of this paper is to consolidate work by these eral, make subtle distinctions clear. Continued analysis of authors on the notion of rigidity in OntoClean. We establish the formalizations themselves has shown that there were a unifying framework that elaborates the dimensions along more distinctions than originally realized, and will be the which the recent proposed accounts of rigidity differ. These subject of this paper. dimensions are alethic modality, time, and existence. We Modal logic was chosen for the formalization mainly due to apply this framework to develop a range of definitions that, the needs of specifying the semantics of rigidity, in particu- given basic intuitions of how time and existence influences lar anti-rigidity. We need to express the fact that some intuitions about rigidity, we claim cover the space of possi- class, e.g. HospitalPatient, is anti-rigid, that any instance of ble definitions. this class must possibly not be an instance. A purely tempo- ral axiomatization would require that every instance does 1 In this paper we uniformly use the term “property” to denote change, which is not true. A person who was a hospital unary relations in intension, as was the meaning of the term in the patient for his entire existence should not contradict the anti- original work on OntoClean done by Guarino and Welty and as is rigidity of the class – it was always possible for him to be- standard usage in philosophical literature. We acknowledge, how- come a non-patient, it just never happened. ever, that, in addition, the terms “class”, “kind”, and sometimes OntoClean was formalized in S5 modal logic with the Bar- “type” are commonly used. When talking about properties that can Formula, which gives us a constant domain (every ob- range over properties, we use the term “meta-property”. ject exists in every possible world) and universal accessibil- which amounts to saying the extension of a temporally rigid ity (every world is accessible from every other world). The property must be the same for all time points and all possi- domain of quantification is possibilia, which when com- ble worlds. bined with S5+BF introduces a need for an actual existence Andersen and Menzel [2004] pointed out that [A6] causes predicate (E), as opposed to logical existence, that indicates problems for non-exemplifiable properties, and does not some object actually exists in the possible world [Miller, accurately capture the intuition expressed as, “An instance 2002]. The unary existence predicate indicates timeless ex- of a rigid property cannot cease to be an instance of that istence, and the binary predicate indicates existence at a property, unless it ceases to exist,” [Welty & Guarino, 2001] particular time in the possible world. since [A6] requires an entity to have the property always Given the separate (i.e. non-modal) treatment of time within and in all possible worlds, e.g. if Person is a rigid class, possible worlds implied by the OntoClean formalizations, must Aristotle be a person even in a possible world in which we need to clarify the intuition that possible worlds have a he does not exist? To address this, they proposed: time line within them, which we also assume to be totally [A7] ±∀xt &(x,t) → ±∀t’(E(x,t’) → &(x,t’)) ordered wrt to a succession relation (<) [Artale & Lutz, (temporally existential rigidity) 2004]. Given the constant domain assumption, times across worlds are the same, and we further assume that, in order to and introduced the constraint (which we have adopted) that be the same, their ordering is maintained across worlds as & is restricted to only exemplifiable properties. Citing many of the same problems, Carrera et al [2004] pro- well: posed: [A1] ∀t t ¡ t < t → ± t < t 1 2 1 2 1 2 [A8] ±∀xt (E(x,t) . &(x,t)) → ±∀t’(E(x,t’) → &(x,t’)) We require a strong notion of subsumption in order to have which is similar to [A7], with a slightly stonger restriction modal consequences for the meta-properties [Kaplan, 2001]: on existence. [A2] subsumes(ψ,&) ↔ ± ∀x &(x) → ψ(x) Both Andersen&Menzel and Carrara et al point out that In order to prevent trivial satisfaction of the axioms, we their accounts of rigidity, by introducing actual existence in require that all properties be exemplifiable [Andersen & the antecedents, say nothing about what happens to entities Menzel, 2004], i.e. for any property &: when they do not exist, leaving open the possibility that an [A3] ¡∃x &(x) instance of a rigid property could change its membership when it does not exist. These assumptions may be somewhat controversial in a phi- losophical setting, however we believe for practical use in ontology engineering, these assumptions are widespread, 4 Kinds of Rigidity though often implicit. One of the original motivations for specifying the Onto- Clean meta-properties was to encourage more rigorous 3 History of Rigidity analysis of the properties in an ontology in order to make the meaning more clear. This clarity would also help to ex- Although not the central notion in OntoClean, Rigidity is pose differences in meaning between ontologies that shared the simplest and most intuitive, with more obvious immedi- terms, e.g. if two ontologies used a class named Person, but ate utility and applicability across the ontology, conceptual it was rigid in one ontology and non-rigid in another, then modeling, and domain modeling communities, and this has we would have some indication that the properties mean made it the most carefully studied of the OntoClean different things. metaproperties. The different accounts of rigidity above ([A5]-[A8]) were The general intuition of rigidity is that certain properties in all intended as corrections on the original formulation, in an an ontology are essential to their instances – an instance effort to come up with the one true formalization of the no- cannot change its membership, and was first axiomatized tion. In this paper, we take a slightly different stance: all of [Guarino & Welty, 2000a] as a property is rigid iff: & the accounts above are correct, and specify different kinds of [A4] ∀x &(x) → ±&(x) rigidity. This has encouraged us to clarify these distinctions Kaplan [2001] pointed out that [A4] is not a straightforward further, in order to establish a more general framework for extension of Kripke’s rigid designators [Kripke, 1982] to rigidity.
Load more

Recommended publications
  • Lecture Notes in Computer Science 7032 Commenced Publication in 1973 Founding and Former Series Editors: Gerhard Goos, Juris Hartmanis, and Jan Van Leeuwen
    Lecture Notes in Computer Science 7032 Commenced Publication in 1973 Founding and Former Series Editors: Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen Editorial Board David Hutchison Lancaster University, UK Takeo Kanade Carnegie Mellon University, Pittsburgh, PA, USA Josef Kittler University of Surrey, Guildford, UK Jon M. Kleinberg Cornell University, Ithaca, NY, USA Alfred Kobsa University of California, Irvine, CA, USA Friedemann Mattern ETH Zurich, Switzerland John C. Mitchell Stanford University, CA, USA Moni Naor Weizmann Institute of Science, Rehovot, Israel Oscar Nierstrasz University of Bern, Switzerland C. Pandu Rangan Indian Institute of Technology, Madras, India Bernhard Steffen TU Dortmund University, Germany Madhu Sudan Microsoft Research, Cambridge, MA, USA Demetri Terzopoulos University of California, Los Angeles, CA, USA Doug Tygar University of California, Berkeley, CA, USA Gerhard Weikum Max Planck Institute for Informatics, Saarbruecken, Germany Lora Aroyo Chris Welty Harith Alani Jamie Taylor Abraham Bernstein Lalana Kagal Natasha Noy Eva Blomqvist (Eds.) The Semantic Web – ISWC 2011 10th International Semantic Web Conference Bonn, Germany, October 23-27, 2011 Proceedings, Part II 13 Volume Editors Lora Aroyo VU University Amsterdam, The Netherlands; [email protected] Chris Welty IBM Research, Yorktown Heights, NY, USA; [email protected] Harith Alani The Open University, Milton Keynes, UK; [email protected] Jamie Taylor Google, Mountain View, CA, USA; [email protected] Abraham Bernstein University of Zurich,
    [Show full text]
  • An Overview of Ontoclean
    8 An Overview of OntoClean Nieola Guarino1 and Christopher A. Welty lLaboratory for Applied Ontology (ISTC-CNR) Polo Tecnologico, Via Solteri 38, 38100 Trento, ITAL Y [email protected] 2IBM Watson Research Center 19 Skyline Dr., Hawthome, NY 10532, USA [email protected] Summary. OntoClean is a methodology for validating the ontologie al adequaey of taxonomie relationships. It is based on highly general ontologieal notions drawn from philosophy, like essence, identity, and unity, whieh are used to ehar­ aeterize relevant aspeets of the intended meaning of the properties, classes, and relations that make up an ontology. These aspeets are represented by formal metaproperties, whieh impose several eonstraints on the taxonomie strueture of an ontology. The analysis of these eonstraints helps in evaluating and validating the ehoiees made. In this ehapter we present an informal overview ofthe philoso­ phieal notions involved and their role in OntoClean, review some eommon onto­ logieal pitfalls, and walk through the example that has appeared in pieees in pre­ vious papers and has been the basis of numerous tutorials and talks. 8.1 Introduction The OntoClean methodology was first introdueed in aseries of eonferenee-Iength papers in 2000 [Guarino and Welty, 2000a-e; Welty and Guarino, 2001], and re­ eeived mueh attention and use in subsequent years. The main eontribution of On­ toClean was the beginning of a formal foundation for ontologieal analysis. Alan Reetor, a seasoned veteran at ontologieal analysis in the medieal domain, said of OntoClean, " ... what you have done is reduee the amount of time I spend arguing with doetors that the way I want to model the world is right..." [Reetor, 2002].
    [Show full text]
  • Cristobalthesis
    Open Personalization: Involving Third Parties in Improving the User Experience of Websites Dissertation presented to the Department of Computer Languages and Systems of the University of the Basque Country in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy (“international” mention) Cristóbal Arellano Bartolomé Supervisors: Prof. Dr. Oscar Díaz García Dr. Jon Iturrioz Sánchez San Sebastián, Spain, 2013 This work was hosted by the University of the Basque Country (Faculty of Computer Sciences). The author enjoyed a doctoral grant under de FPI (Formacion de Personal Investigador) from the Spanish Ministry of Science & Education during the years 2007 to 2011. The work was was co- supported by the Spanish Ministry of Education, and the European Social Fund under contracts (TIN2005-05610), MODELINE (TIN2008-06507- C02-01) and Scriptongue (TIN2011-23839). Summary Traditional software development captures the user needs during the requirement analysis. The Web makes this endeavour even harder due to the difficulty to determine who these users are. In an attempt to tackle the heterogeneity of the user base, Web Personalization techniques are proposed to guide the users’ experience. In addition, Open Innovation allows organisations to look beyond their internal resources to develop new products or improve existing processes. This thesis sits in between by introducing Open Personalization as a means to incorporate actors other than webmasters in the personalization of web applications. The aim is to provide the technological basis that builds up a trusty environment for webmasters and companion actors to collaborate, i.e. "an architecture of participation". Such architecture very much depends on these actors’ profile.
    [Show full text]
  • Semantic Web: a Review of the Field Pascal Hitzler [email protected] Kansas State University Manhattan, Kansas, USA
    Semantic Web: A Review Of The Field Pascal Hitzler [email protected] Kansas State University Manhattan, Kansas, USA ABSTRACT which would probably produce a rather different narrative of the We review two decades of Semantic Web research and applica- history and the current state of the art of the field. I therefore do tions, discuss relationships to some other disciplines, and current not strive to achieve the impossible task of presenting something challenges in the field. close to a consensus – such a thing seems still elusive. However I do point out here, and sometimes within the narrative, that there CCS CONCEPTS are a good number of alternative perspectives. The review is also necessarily very selective, because Semantic • Information systems → Graph-based database models; In- Web is a rich field of diverse research and applications, borrowing formation integration; Semantic web description languages; from many disciplines within or adjacent to computer science, Ontologies; • Computing methodologies → Description log- and a brief review like this one cannot possibly be exhaustive or ics; Ontology engineering. give due credit to all important individual contributions. I do hope KEYWORDS that I have captured what many would consider key areas of the Semantic Web field. For the reader interested in obtaining amore Semantic Web, ontology, knowledge graph, linked data detailed overview, I recommend perusing the major publication ACM Reference Format: outlets in the field: The Semantic Web journal,1 the Journal of Pascal Hitzler. 2020. Semantic Web: A Review Of The Field. In Proceedings Web Semantics,2 and the proceedings of the annual International of . ACM, New York, NY, USA, 7 pages.
    [Show full text]
  • Introduction to Semantic Web Technologies & Linked Data
    IntroductionIntroduction toto SemanticSemantic WebWeb TechnologiesTechnologies && LinkedLinked DataData OktieOktie HassanzadehHassanzadeh UniversityUniversity ofof TorontoToronto March 2011 CS 443: Database Management Systems - Winter 2011 Outline 2 Introduction Semantic Web Technologies Resource Description Framework (RDF) Querying RDF data (SPARQL) Linked Data Linked Data Principles Linking Open Data Community Project Example Data Sources Example Applications 3 Introduction Web of Documents vs. Web of Data Web of Documents 4 Untyped Untyped Untyped Links Links Links API/ HTML HTML HTML XML A B C D Primary objects: documents Links between documents (or parts of them) Degree of structure in data: fairly low Implicit semantics of contents Designed for: human consumption Based on presentations by Chris Bizer, Richard Cyganiak, Tom Heath, available at http://linkeddata.org/guides-and-tutorials Web of Documents: Problem 5 ? thing thing Are two documents talking about the same ? ? “thing”? ? ? ? Untyped Untyped Untyped Links Links Links API/ HTML HTML HTML XML A B C D Based on presentations by Chris Bizer, Richard Cyganiak, Tom Heath, available at http://linkeddata.org/guides-and-tutorials Example Query 6 Elvis Presley 1935 - 1977 Will there ever be someone like him again? Based on presentation by Lauw, Schenkel, Suchanek, Theobald and Weikum, available at http://www.mpi-inf.mpg.de/yago-naga/CIKM10-tutorial/ Example Query 7 Another Elvis Elvis Presley: The Early Years Elvis spent more weeks at the top of the charts than any other artist. www.fiftiesweb.com/elvis.htm Based on presentation by Lauw, Schenkel, Suchanek, Theobald and Weikum, available at http://www.mpi-inf.mpg.de/yago-naga/CIKM10-tutorial/ Example Query 8 Another singer called Elvis, young Personal relationships of Elvis Presley – Wikipedia ...when Elvis was a young teen...
    [Show full text]
  • Semantics Driven Human-Machine Computation Framework for Linked Islamic Knowledge Engineering
    Semantics Driven Human-Machine Computation Framework for Linked Islamic Knowledge Engineering by Amna Basharat (Under the Direction of Khaled Rasheed and I. Budak Arpinar) Abstract Formalized knowledge engineering activities including semantic annotation and linked data management tasks in specialized domains suffer from considerable knowledge acquisition bottleneck - owing to the lack of availability of experts and in-efficacy of automated approaches. Human Computation & Crowdsourcing (HC&C) methods successfully advocate leveraging the human intelligence and pro- cessing power to solve problems that are still difficult to be solved computationally. Contextualized to the domain of Islamic Knowledge, this research investigates the synergistic interplay of these HC&C methods and the semantic web and proposes a semantics driven human-machine computation framework for knowledge engineer- ing in specialized and knowledge intensive domains. The overall objective is to augment the process of automated knowledge extraction and text mining methods using a hybrid approach for combining collective intelligence of the crowds with that of experts to facilitate activities in formalized knowledge engineering - thus overcoming the so-called knowledge acquisition bottleneck. As part of this framework, we design and implement formal and scalable knowl- edge acquisition workflows through the application of semantics driven crowdsourc- ing methodology and its specialized derivative, called learnersourcing. We evaluate these methods and workflows for a range of knowledge engineering tasks including thematic classification, thematic disambiguation, thematic annotation and contex- tual interlinking for two primary Islamic texts, namely the Qur’an and the books of Prophetic narrations called the Hadith. This is done at various levels of granu- larity, including atomic and composite task workflows, that existing research fails to address.
    [Show full text]
  • An Overview of Ontoclean
    8 An Overview of OntoClean Nicola Guarino1 and Christopher A. Welty2 1Laboratory for Applied Ontology (ISTC-CNR) Polo Tecnologico, Via Solteri 38, 38100 Trento, ITALY [email protected] 2IBM Watson Research Center 19 Skyline Dr., Hawthorne, NY 10532, USA [email protected] Summary. OntoClean is a methodology for validating the ontological adequacy of taxonomic relationships. It is based on highly general ontological notions drawn from philosophy, like essence, identity, and unity, which are used to characterize relevant aspects of the intended meaning of the properties, classes, and relations that make up an ontology. These aspects are represented by formal metaproperties, which impose several constraints on the taxonomic structure of an ontology. The analysis of these constraints helps in evaluating and validating the choices made. In this chapter we present an informal overview of the philosophical notions in- volved and their role in OntoClean, review some common ontological pitfalls, and walk through the example that has appeared in pieces in previous papers and has been the basis of numerous tutorials and talks. 8.1 Introduction The OntoClean methodology was first introduced in a series of conference-length papers in 2000 [Guarino and Welty, 2000a-c; Welty and Guarino, 2001], and re- ceived much attention and use in subsequent years. The main contribution of On- toClean was the beginning of a formal foundation for ontological analysis. Alan Rector, a seasoned veteran at ontological analysis in the medical domain, said of OntoClean, “…what you have done is reduce the amount of time I spend arguing with doctors that the way I want to model the world is right…” [Rector, 2002].
    [Show full text]
  • Ontology-Driven Conceptual Modeling
    Ontology-Driven Conceptual Modeling Chris Welty IBM Watson Research Center CAISE-02 Acknowledgements People Organizations Vassar College, USA Nicola Guarino LADSEB-CNR, Padova Cladio Masolo CNR Cognitive Science Aldo Gangemi Institute, Trento Alessandro Oltramari Bill Andersen OntologyWorks, Inc. CAISE-02 2 1 Outline • Setting the record straight • Motivation • Formal foundation • “Upper Level” distinctions • Common pitfalls CAISE-02 3 What is Ontology? • A discipline of Philosophy – Meta-physics dates back to Artistotle • Meta (after) + physica (physical, real) – Ontology dates back to 17th century • Ontos (that which exists) + logos (knowledge of) • As in TorONTO, ONTario, ON TOp – The science of what is (in the universe) – “One universe, One ontology” • Quine, 1969: “To exist is to be the value of a quantified variable” CAISE-02 4 2 What is Ontology? • Borrowed by AI community – McCarthy (1980) calls for “a list of things that exist” – Specify all the kinds of things that can be the values of variables • Evolution of meaning in CS – Now refers to domain modeling, conceptual modeling, knowledge engineering, etc. • Note: not a “new name for an old thing” CAISE-02 5 What is an Ontology? • Poor definition: “Specification of a conceptualization” [Gruber, 1993] • Better: “Description of the kinds of entities there are and how they are related.” • Good ontologies should provide: – Meaning –Agreement – Organization – Common Understanding – Taxonomy – Vocabulary – Connection to the “real world” CAISE-02 6 3 What is an Ontology? a set of a collection
    [Show full text]
  • IBM Research Report Harnessing Disagreement in Crowdsourcing A
    RC25371 (WAT1304-058) April 23, 2013 Computer Science IBM Research Report Harnessing Disagreement in Crowdsourcing a Relation Extraction Gold Standard Lora Aroyo VU University Amsterdam, The Netherlands Chris Welty IBM Research Division Thomas J. Watson Research Center P.O. Box 208 Yorktown Heights, NY 10598 USA Research Division Almaden - Austin - Beijing - Cambridge - Haifa - India - T. J. Watson - Tokyo - Zurich LIMITED DISTRIBUTION NOTICE: This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. It has been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g. , payment of royalties). Copies may be requested from IBM T. J. Watson Research Center , P. O. Box 218, Yorktown Heights, NY 10598 USA (email: [email protected]). Some reports are available on the internet at http://domino.watson.ibm.com/library/CyberDig.nsf/home . Harnessing disagreement in crowdsourcing a relation extraction gold standard Lora Aroyo Chris Welty VU University Amsterdam IBM Watson Research Center The Netherlands USA [email protected] [email protected] ABSTRACT are described in natural language. The importance of rela- One of the first steps in any kind of web data analytics tions and their interpretation is widely recognized in NLP, is creating a human annotated gold standard. These gold but whereas NLP technology for detecting entities (such as standards are created based on the assumption that for each people, places, organizations, etc.) in text can be expected annotated instance there is a single right answer.
    [Show full text]
  • Crowdsourcing,The,Semantic,Web
    ! ! ! ! ! ! Proceedings! of!the! First!International!Workshop! on! ! Crowdsourcing,the,Semantic,Web, held%in%conjunction%with! the!12th!International!Semantic!Web!Conference! ! Sydney,!Australia! ! Editors:! Maribel!Acosta! Lora!Aroyo! Abraham!Bernstein! Jens!Lehmann! Natasha!Noy! Elena!Simperl! ! Copyright!©!2013!for!the!individual!papers!by!the!papers’!authors.!Copying! permitted!only!for!private!and!academic!purposes.!This!volume!is!published!and! copyrighted!by!its!editors.! Preface This volume contains the papers presented at the 1st International Workshop on ”Crowdsourcing the Semantic Web” that was held in conjunction with the 12th International Semantic Web Conference (ISWC 2013), 21-25 October 2013, in Sydney, Australia. This interactive workshop takes stock of the emergent work and chart the research agenda with interactive sessions to brainstorm ideas and potential applications of collective intelligence to solving AI hard semantic web problems. There were 12 submissions. Each submission was reviewed by at least 2, and on the average 3, program committee members. The committee decided to accept 9 papers. Our special thanks goes to the reviewers who diligently reviewed the papers within this volume. September 3, 2013 Maribel Acosta Lora Aroyo Abraham Bernstein Jens Lehmann Natasha Noy Elena Simperl v Table of Contents Full Papers Crowdsourced Semantics with Semantic Tagging: “Don’t just tag it, LexiTag it!” ...................................................... 1 Csaba Veres ”Dr. Detective”: combining gamification techniques and crowdsourcing to create a gold standard in medical text ............................. 16 Anca Dumitrache, Lora Aroyo, Chris Welty, Robert-Jan Sips and An- thony Levas SLUA: Towards Semantic Linking of Users with Actions in Crowdsourcing 32 Umair Ul Hassan, Sean O’Riain and Edward Curry Content and Behaviour Based Metrics for Crowd Truth ...............
    [Show full text]
  • The Disagreement Deconvolution: Bringing Machine Learning Performance Metrics in Line with Reality
    The Disagreement Deconvolution: Bringing Machine Learning Performance Metrics In Line With Reality Mitchell L. Gordon Kaitlyn Zhou Kayur Patel Stanford University Stanford University Apple Inc. [email protected] [email protected] [email protected] Tatsunori Hashimoto Michael S. Bernstein Stanford University Stanford University [email protected] [email protected] ABSTRACT 1 INTRODUCTION Machine learning classifiers for human-facing tasks such as com- The machine learning classifiers that underpin modern social com- ment toxicity and misinformation often score highly on metrics puting systems such as Facebook, Wikipedia, and Twitter are a such as ROC AUC but are received poorly in practice. Why this gap? study in contrasts. On one hand, current performance metrics for Today, metrics such as ROC AUC, precision, and recall are used to popular tasks such as comment toxicity classification and disin- measure technical performance; however, human-computer inter- formation detection are extremely high, featuring up to .95 ROC action observes that evaluation of human-facing systems should AUC [11, 72], making the problem appear nearly solved. On the account for people’s reactions to the system. In this paper, we intro- other hand, audits suggest that these algorithms’ performance is duce a transformation that more closely aligns machine learning in reality much poorer than advertised [56]. Even well-intentioned classification metrics with the values and methods of user-facing platforms will continue to make highly-publicized mistakes [16] performance measures. The disagreement deconvolution takes in until they can better align metrics with reality. any multi-annotator (e.g., crowdsourced) dataset, disentangles sta- This disconnect between classifier performance and user-facing ble opinions from noise by estimating intra-annotator consistency, performance is indicative of a larger disconnect between how ma- and compares each test set prediction to the individual stable opin- chine learning (ML) and human-computer interaction (HCI) re- ions from each annotator.
    [Show full text]
  • Evaluating Ontology Cleaning
    Evaluating Ontology Cleaning Christopher Welty, Ruchi Mahindru, and Jennifer Chu-Carroll IBM T.J. Watson Research Center 19 Skyline Dr. Hawthorne, NY 10532 {welty,jencc}@us.ibm.com, [email protected] Abstract very little evidence that it can have impact on Ontology as a discipline of Computer Science has made knowledge-based systems. In fact, there appears many claims about its usefulness, however to date there has to be a significant obstacle in understanding the been very little evaluation of those claims. We present the methodology, and even without this “learning results of an experiment using a hybrid search system with a curve”, significant manual effort must be significant knowledge-based component to measure, using expended to employ the methodology to develop precision and recall, the impact of improving the quality of actual “clean” ontologies. Furthermore, there an ontology on overall performance. We demonstrate that has been no clear argument that such an improving the ontology using OntoClean (Guarino and expenditure will pay for itself in the long run. Welty, 2002), does positively impact performance, and that Indeed, “Why does it matter?” has been the most having knowledge of the search domain is more effective than domain-knowledge-free search techniques such as link frequent criticism of the OntoClean approach. analysis. We report here on a series of experiments using a Content Areas: Ontologies hybrid search system with a significant knowledge-based component to test the impact of improving the quality of ontologies on system Introduction performance. The use of search as a test system provides a well understood framework for Ontologies have been proposed as a solution to empirical evaluation, and gives an excellent numerous problems in areas such as search, opportunity to address the “Why does it matter?” semantic integration, agents, human-computer question.
    [Show full text]