Ontologies. Course : Software Agent Technology Santtu Toivonen, VTT Information Technology

Transcription

1 ntologies Course : Software gent Technology Santtu Toivonen, VTT Information Technology

2 VTT INFRMTIN TECHNLGY bout me Work Research Scientist VTT Information Technology Education Master of rts University of Helsinki, major in Cognitive Science Interests gent communication ntologies for agents Distribution of cognition Philosophy of mind

3 VTT INFRMTIN TECHNLGY utline What is an ontology? What is a concept? Some simple sample ontologies Size and scope of an ontology ntology management The Semantic Web Some ontology representation languages and tools Further material

4 VTT INFRMTIN TECHNLGY ntology hmmm, sounds familiar You have come across ontologies before in this course: (cfp :sender (agent-identifier :name j) :receiver (set (agent-identifier :name i)) :content "((action (agent-identifier :name i) (sell plum 50)) (any?x (and (= (price plum)?x) (<?x 10))))" :ontology fruit-market :language fipa-sl) but what are they?

5 VTT INFRMTIN TECHNLGY What is an ontology? Webster: Main Entry: on tol o gy Pronunciation: än-'tä-l&-je Function: noun Etymology: New Latin ontologia, from ont- + -logia -logy Date: circa : a branch of metaphysics concerned with the nature and relations of being 2 : a particular theory about the nature of being or the kinds of existents

6 VTT INFRMTIN TECHNLGY What is an ontology? Tom Gruber: Short answer: n ontology is a specification of a conceptualization. [ ] an ontology is a description (like a formal specification of a program) of the concepts and relationships that can exist for an agent or a community of agents.

7 VTT INFRMTIN TECHNLGY What is an ontology? Nicola Guarino: [ ] in I, an ontology refers to an engineering artifact, constituted by a specific vocabulary used to describe a certain reality, plus a set of explicit assumptions regarding the intended meaning of the vocabulary words.

8 VTT INFRMTIN TECHNLGY ntologies vs. data models No strict line in between, but ontologies are more general more reusable intended for multiple purposes, goals, and users more easily shareable take stand on semantics of concepts (as opposed to mere structure and integrity)

9 VTT INFRMTIN TECHNLGY What is a concept? Concepts (among other things) are in general language-independent (words 'cat' and 'kissa' denote the same concept) are mental or logical representations of reality are related to other concepts do not need symbols but hold them for means of communication concept has intension or meaning extension, i.e. the set of objects that the concept refers to n the difference between intension and extension, consider phrases "Evening star" and "Morning star" that have different meanings (intension) yet both refer to planet Venus (extension) ntology is mainly concerned with intension

10 VTT INFRMTIN TECHNLGY "Semiotic Triangle" (or one version of it) Intension Extension 'Cat' Sign or term

11 VTT INFRMTIN TECHNLGY "Semiotic Triangle" (contd.) Intension Extension 'Frank' Sign or term

12 VTT INFRMTIN TECHNLGY "Semiotic Triangle" (contd.) Intension Extension 'Frank is a cat' Sign or term

13 VTT INFRMTIN TECHNLGY simple example ontology Vehicle subclassf Car Boat

14 VTT INFRMTIN TECHNLGY lmost as simple Vehicle subclassf Car ispartf Wheel

15 VTT INFRMTIN TECHNLGY Gets a bit tricky... Vehicle type Class Property type type subclassf type Car ispartf Wheel type

16 VTT INFRMTIN TECHNLGY subclassf Thing relations multiply easily Vehicle type Class Property type type subclassf type Car ispartf Wheel type

17 VTT INFRMTIN TECHNLGY Important concepts wrt. ontology Concepts abstract or concrete, elementary or composite, real or fictitious can have attributes (properties) Taxonomy (typically at least subclass-relation) ther relations part of, connected to, on top of, unary vs. binary vs. n-ary relations xioms (sentences that are always true) axioms typically built using logic (first or higher order) Instances, Individuals, Facts, Claims extension Production rules and inference mechanisms

18 VTT INFRMTIN TECHNLGY n aside: "Type vs. subclass confusion" In general: type-relation connects an instance with the class it is an instance of subclass-relation connects a class with its superclass nimal subclassf Mammal subclassf type Human

19 VTT INFRMTIN TECHNLGY "Type vs. subclass confusion" (contd.) The confusion: Which relation to choose with more abstract or complex concepts? Paying Phone Paying with VIS subclassf / type? Nokia 6100 subclassf / type? This problem emerges when chaining ontologies (i.e. defining ontologies for ontologies)

20 VTT INFRMTIN TECHNLGY Size and scope of an ontology Two extremes (the reality something in between): ne huge ontology that captures "everything" ne (small) ontology for each specific application

21 VTT INFRMTIN TECHNLGY "ne large ontology" approach Benefits few or no internal inconsistencies for an application developer easier to find uniform documentation less overlapping work! Drawbacks who maintains it? who is responsible? heavy and slow to use (both for human user and for application) difficult to take into account everybody's opinions and wishes Example: Cyc

22 VTT INFRMTIN TECHNLGY "Several small ontologies" approach Benefits ontologies fit well the application demands faster to use easier to form complete picture of an ontology (fewer concepts and interrelations) Drawbacks different ontologies do not fit together without either central coordination body or ontology mapping software M overlapping work - same concepts defined in multiple ontologies, either in the same way or (even worse!) differently

23 VTT INFRMTIN TECHNLGY Some mixed / other approaches upper ontology domains M mediating software none

24 VTT INFRMTIN TECHNLGY Which of the approaches to use with agent systems (in the long run)? gent-systems are typically distributed systems I.e. agent-systems consisting of but one agent are rare agents typically need to communicate with each other agents should understand each other People often design and implement agents independently of each other agents' understanding each other is (partly) based on ontologies M

25 VTT INFRMTIN TECHNLGY ntology management Individuals and instances change over time, whereas ontologies are intended to be relatively stable and static but ontology management is needed Issues on interdependencies upper ontology domain ontologies domain ontology domain ontology ontology applications pproaches for cross-usage of concepts defined in multiple ontologies integration (take existing ontologies and build a new one) merge (unify existing ontologies) mapping software ( runtime approach) M

26 VTT INFRMTIN TECHNLGY ntology management (contd.) ntology versioning an important aspect Five ways for an ontology to change 1. Changes in the conceptual hierarchy (logical changes) 2. Changes in the documentation of the concepts (non-logical changes) 3. Renaming concepts (identifier changes) 4. Defining new concepts 5. Deleting concepts Prospective usage: using data sources that conform with the previous version of the ontology via the new version Retrospective usage: using data sources that conform with the new version of the ontology via the previous version

27 VTT INFRMTIN TECHNLGY ntology management (contd.) Compatibility types 1. Complete compatible revisions the semantics of the ontology is not changed, for example, syntactic changes or updates of natural language descriptions; this type of change is compatible in both prospective and retrospective use 2. Backward compatible revisions the semantics of the ontology are changed in such a way that the interpretation of data via the new ontology is the same as when using the previous version of the ontology, for example, the addition of an independent class; this type of change is compatible in prospective use 3. Upward compatible revisions the semantics of the ontology is changed is such a way that an older version can be used to interpret newer data sources correctly, for example, the removal of an independent class; this revision is compatible in retrospective use 4. Incompatible revisions the semantics of the ontology is changed in such a way that the interpretation of old data sources is invalid, for example, changing the place in the hierarchy of a class; this type of change is incompatible in both prospective use and retrospective use

28 VTT INFRMTIN TECHNLGY The case of Semantic Web Gathers a lot of attention at the moment ontology as a buzzword "Next generation of the web", "intelligent web", "autonomous web", Core idea: "The Semantic Web is not a separate Web but an extension of the current one, in which information is given well-defined meaning, better enabling computers and people to work in cooperation." From Berners-Lee et al., Semantic Web vs. WWW as it is at the moment (emphasis on layout and structure of (HTML etc.) pages) Machine-understandable content of Semantic Web is defined in ontologies Relationship and relevance to this course: machine-processing of Semantic Web content likely to happen via agents (or something alike) W3C has recently ( ) launched Phase 2 of Semantic Web ctivity Includes a new Best Practices and Deployment WG

29 VTT INFRMTIN TECHNLGY The "layer cake" of Semantic Web Reprinted from a presentation by Tim Berners-Lee in XML2000 Conference, URL:

30 VTT INFRMTIN TECHNLGY Technologies of Semantic Web RDF (Resource Description Framework) Triples: Subject, Predicate, bject RDFS (RDF Schema) Vocabulary definitions for RDF (cf. DTD and XML Schema for XML) subclassf, type, domain, range,... DML (DRP gent Markup Language) IL (ntology Inference Layer or ntology Interchange Language) WL (Web ntology Language) DML, IL DML+IL WL US W3C Europe

31 VTT INFRMTIN TECHNLGY WL WL builds on the work of DML+IL DML+IL an extension of RDF Schema W3C Recommendation since Inconsistencies with RDF(S) and DML+IL problems with the layered approach presented earlier three versions of WL: WL Full Union of WL and RDF(S) WL DL (Description Logics) Restricted to DL/FL fragment ( DML+IL) WL Lite Subset of WL DL

32 VTT INFRMTIN TECHNLGY WL-S WL Services is a language for describing Web Services cronyms often mentioned as Web Services technologies: WSDL, Web Services Description Language SP, Simple bject ccess Protocol UDDI, Universal Description, Discovery and Integration However, these technologies are insufficient for describing the semantics of Web Services in a machine-understandable manner (what the services do and how they do it) enter WL-S Builds on WL, currently version 1.0 (November 2003) (draft version 1.1 also online, Jan 2004) Possible to map WL-S to WSDL nkolekar et al. (2002)

33 VTT INFRMTIN TECHNLGY Some other ontology representation languages "Traditional" ontology representation languages (not web based) ntolingua (based on KIF) KBC, pen Knowledge Base Connectivity CML, perational Conceptual Modeling Language LM FLogic, Frame Logic Web based ontology representation languages XL, XML-Based ntology Exchange Language SHE, Simple HTML ntology Extension

34 VTT INFRMTIN TECHNLGY Some ontology representation languages compared Concepts Relations Functions Procedures Instances xioms Production rules ntol KBC CML LM FLogic XL SHE RDF(S) IL / / / / / /- "traditional" web based

35 VTT INFRMTIN TECHNLGY Summary of ontology representation language comparison For interchanging ontologies on the web, web based languages are a better solution For representing or modeling ontologies, "traditional" languages are a better solution as far as the expressiveness is concerned However, XML based languages function well for taxonomies with a relatively simple structure For performing reasoning, "traditional" languages are a better choice Source: Gorcho and Gómez-Pérez (2000) Note: some languages have developed since the writing of the article (e.g. IL DML+IL WL)

36 VTT INFRMTIN TECHNLGY The case of Cyc Cycorp: "It's just common sense" In CycL over hand-entered assertions (or "rules") designed to capture a large portion of what we normally consider consensus knowledge about the world ("river flows downhill" etc.) Sometimes so called microtheories are however necessary difference with domain ontologies? Nevertheless exemplifies the "ne large ontology" approach! pencyc: an open source subset of Cyc concepts: an upper ontology for all of human consensus reality assertions about the concepts, interrelating them, constraining them, in effect (partially) defining them. pencyc has DML serialization Note however that Cyc is not a Semantic Web phenomenon but been under development since 1984 There also exists CycL-to-Lisp, CycL-to-C, etc. translators

37 VTT INFRMTIN TECHNLGY Standard Upper ntology (SU) IEEE: SUM (Suggested Upper Merged ntology) Serializations in KIF, DML, LM Web based ontology browsers SU IFF (Information Flow Framework) IFF provides the terminology, semantics and principled foundation for a metalevel ontological framework a framework for sharing ontologies, manipulating ontologies as objects, relating ontologies through morphisms, partitioning ontologies, composing ontologies via colimits, discussing ontological structure, noting dependencies between ontologies, declaring the use of other ontologies, etc

38 VTT INFRMTIN TECHNLGY Some ontology tools ntolingua collaborative environment to browse, create, edit, modify, and use ontologies Protégé (ntology editor) Lots of plugins: KBC, RDF(S), WL, UML, XML, XML Schema, Prolog, Jess, etc. WL-S to appear iled (ntology editor for IL) Jena Java PI for manipulating RDF, RDF(S), and WL

39 VTT INFRMTIN TECHNLGY The role of ontologies in agent communication Used by agents when trying to make sense of each other Case of FIP: ntology is denoted by one parameter in an CL message (along with other parameters: sender, receiver, content, reply-with, replyby, in-reply-to, envelope, language, protocol, and conversation-id) Example: (cfp :sender (agent-identifier :name j) :receiver (set (agent-identifier :name i)) :content "((action (agent-identifier :name i) (sell plum 50)) (any?x (and (= (price plum)?x) (<?x 10))))" :ontology fruit-market :language fipa-sl)

40 VTT INFRMTIN TECHNLGY The role of ontologies in agent communication (contd.) ntology is referenced from within CL message and it determines the semantics of the concepts used in the content language However, ontologies themselves might concern whatever topics ne interesting case would be to "ontologize" agent communication For example interaction protocols (Conversation layer) might be described in an ontology external to the agents The agents could download the interaction protocol descriptions and modify their behavior accordingly ntology of actions or tasks vs. ontology of facts Know-how vs. know-that

41 VTT INFRMTIN TECHNLGY The role of ontologies in agent communication (contd.)

42 VTT INFRMTIN TECHNLGY Further material rticles, presentations and books Thomas R. Gruber (1993): Translation pproach to Portable ntology Specifications. Knowledge cquisition, 5(2): Thomas R. Gruber (1993): Toward principles for the design of ontologies used for knowledge sharing. llen Newell (1982): The Knowledge Level. rtificial Intelligence 18(1): nkolekar et al. (2002): DML-S: Web Service Description for the Semantic Web. Proceedings of the First International Semantic Web Conference (ISWC2002) Heidelberg, Germany: Springer- Verlag.

43 VTT INFRMTIN TECHNLGY Further material (contd.) rticles, presentations and books (contd.) Gottlob Frege (1892): Über Sinn und Bedeutung. In: Zeitschrift für Philosophie und philosophische Kritik 100: Meersman et al. (2002): Data modelling versus ntology engineering. In CM SIGMD Record, December John F. Sowa (2000): Knowledge Representation. Pacific Grove, C: Brooks/Cole. Tim Berners-Lee, James Hendler, and ra Lassila (2001): The Semantic Web. In Scientific merican, May

44 VTT INFRMTIN TECHNLGY Further material (contd.) rticles, presentations and books (contd.) James Hendler (2001): gents and the Semantic Web. In IEEE Intelligent Systems, (16:2): Ian Horrocks (2002): WL verview, ntoweb SIG2 meeting, Innsbruck, ustria, December 17th PoverPoint slides: scar Gorcho and sunción Gómez-Pérez (2000): Evaluating knowledge representation and reasoning capabilities of ontology specification languages. In Proceedings of the ECI 2000 Workshop on pplications of ntologies and Problem-Solving Methods, Berlin, H. Sofia Pinto,. Gómez-Pérez, J. P. Martins. Some Issues on ntology Integration. In Proceedings of IJCI99's Workshop on ntologies and Problem Solving Methods: Lessons Learned and Future Trends,

45 VTT INFRMTIN TECHNLGY Further material (contd.) rticles, presentations and books (contd.) Nicola Guarino (1998): Formal ntology and Information Systems. In Formal ntology in Information Systems. Proceedings of FIS 98, Trento, Italy, 6-8 June msterdam, IS Press, pp Nicola Guarino, Christopher Welty, and Christopher Partridge (2000): Towards ntology-based Harmonization of Web Content Standards. In Proceedings of Conceptual Modeling for E-Business and the Web, ER 2000 Workshops on Conceptual Modeling pproaches for E-Business and The World Wide Web and Conceptual Modeling, Salt Lake City, Utah, US, ctober 9-12, Springer LNCS 1921, pp M. Klein and D. Fensel. ntology Versioning on the Semantic Web. In Proceedings of the International Semantic Web Working Symposium (SWWS), Stanford University, California, US, July,

46 VTT INFRMTIN TECHNLGY Further material (contd.) ctivities, initiatives, web sites, etc. CYC: SU: ntoweb: Semantic Web community portal: W3C Semantic Web activity: WL: ntolingua: KIF: Laboratory for pplied ntology in the Institute of Cognitive Sciences and Technology (ISTC) of Italian National Research Council: