The document discusses developing ontologies, including: 1. What an ontology is and different types of ontologies such as taxonomies, thesauri, and reference models. 2. Representing ontologies using knowledge representation formalisms that have evolved from semantic networks to description logics. 3. The Semantic Web ontology language OWL, which extends RDFS and is divided into three species with different levels of expressivity.

1. Developing Ontologies Joost Breuker

2. overview What is an ontology? Ontology, thesauri and other taxonomic species Core ontologies Representing ontologies (Rinke) Knowledge representation: from T-Boxes to OWL KR tools Using ontologies traditional roles and the Semantic Web tools: reasoning, information retrieval, knowledge management Developing LRI-Core principles of common sense main divisions (phyisical, abstract, mental, role, occurrence)

3. when you miss the points… W3C documentation on semantic web, RDF & OWL http://www.w3.org/2001/sw/ Grigoris Antoniou and Frank van Harmelen. A Semantic Web Primer . MIT-Press, 2004. S. Staab and R. Studer, Handbook on Ontologies , Springer, 2003 F. Baader, et al, (Eds), Description Logic Handbook , Cambridge University Press,2002. (Ch 1, and part III, applications (…Ch 13)

4. Leibniz (1647-1716) on computable ontologies “Once the characteristic numbers of most notions are determined, the human race will have a new kind of tool, a tool that will increase the power of the mind much more than optical lenses helped our eyes, a tool that will be as far superior to microscopes or telescopes as reason is to vision” from: Philosophical Essays

5. Leibniz (1647-1716) on computable ontologies “Once the characteristic numbers of most notions are determined, the human race will have a new kind of tool, a tool that will increase the power of the mind much more than optical lenses helped our eyes, a tool that will be as far superior to microscopes or telescopes as reason is to vision” from: Philosophical Essays computable index concepts reasoning by “ ars combinatorix”

6. what is an ontology? `formal specification of conceptualization’ (Gruber 94)  “ An ontology defines the terms used to describe and represent an area of knowledge ” ( Jeff Heflin, OWL-Use cases, http://www.w3.org/TR/2004/REC-webont-req-20040210/ ) terms: concept (= meaning)… (+ symbol (word; index;…)?) knowledge representation: from informal (eg text) to machine interpretable (via formalization) ontology: `what is’ ≈ what we know

7. an example: newspaper ontology

8. epistemology vs ontology ontology in philosophy: study of existence of entities epistemology: how do we know? epistemology is about justification of knowledge `correct’ reasoning in ontological engineering: ontology: definitional structure of concepts

9. different things or point of view? reasoning method PSM inference deduction abduction classification cover & differentiate PSM hypothesis testing assemble hypothesis match data hypothesis generation predict values obtain data `epistemological’ view `ontological’ view IS-A inferences inferences DEPENDENCY PART-OF

10. (DAML)OWL-S: an `ontology’ for web services

11. Another pseudo-ontology: FOLaw normative reasoning (Valente, Breuker & Brouwer, 99) CASE

12. Is that a problem? Yes : they are reasoning frames by representing reasoning dependencies between types of knowledge/partitions of knowledge bases; not classes (concept definitions) No : OWL (and other KR formalisms) can express easily these frames IMPORTANT: Highly useful in reuse (eg specifying web-services by OWL-S) However: better keep these `epistemological frameworks’ separate

13. Semantic levels of ontology Level 0: Dictionaries , describing informal definitions associated to concept names, with no formal semantic primitives; Level 1: Taxonomies , describing specialization relationships between concepts; Level 2: Thesauri , adding to taxonomies various lexical relationships (hyperonimy, synonimy, partonomy, etc…) eg Wordnet these enable some identification of terms (text) Level 3: Reference models , combining many of the relations above and many other (axiomatic) relations: these enable reasoning…

14. Abstraction levels of ontologies Upper, top, foundational ontologies capturing our most abstract, often common-sense, notions full ontologies (CyC, SUMO, DOLCE, Sowa, LRI-Core…) partial ontologies about: time, space, liquids, physical processes,…

15. Sowa’s (1999) top ontology

16. Abstraction levels of ontologies Upper, top, foundational ontologies capturing our most abstract, often common-sense, notions full ontologies (CyC, SUMO, DOLCE, Sowa, …) partial ontologies about: time, space, liquids, physical processes,… Core ontologies capturing the most abstract terms in a field of practice eg. electro-mechanical engineering, medicine, law, process-industry-components, cultural heritage, etc often: need for including/starting with some `top’ ontology

17. LRI- core ontology for law

18. Abstraction levels of ontologies Upper, top, foundational ontologies capturing our most abstract, often common-sense, notions full ontologies (CyC, SUMO, DOLCE, Sowa, …) partial ontologies about: time, space, liquids, physical processes,… Core ontologies capturing the most abstract terms in a field of practice eg. electro-mechanical engineering, medicine, law, process-industry-components, cultural heritage…---> organic chemistry often: need for including/starting with some `top’ ontology Domain ontologies the `real’ stuff: wines, newspapers, Dutch criminal law, ships,

19. Representing ontologies: a short overview of KR-research core business of AI research two major formalisms: networks-of-concepts and rules from semantic networks (…RDF(S)) to description logics (DL) based systems (KL-ONE `family’) from informal, intuitive `semantics’ to logic based (model theory) trade-off between expressiveness of formalism and tractability of implied inferences distinction between `terminology’ and `assertions’ every country has a capital (generic concepts; T-Box) Amsterdam is the capital of the Netherlands (individuals; A-Box) AI/strict-ontological engineering view: ontology is T-Box in DB/OO community: individuals + schema

20. KR for the Semantic Web: OWL To allow semantics based services and information management, the Web needs protocols and standards that enable: specification, access, and maintenance of the meaning of terms and objects (images) of web-pages also: non-human agents to process pages by their content see http://www.w3.org/TR/2004/REC-webont-req-20040210/ for use-cases Meaning of terms/(images of) objects can be specified in an ontology. Need for an ontology language: RDF(S)?

21. layers of the Semantic Web OWL

22. OWL and RDF(S) RDF(S) knows about: classes (concepts), properties (relations), individuals (instances) but: intractable reasoning allowed (too expressive) lacking expressivity (eg. cardinality, disjunction, properties of properties) Solution: OWL design requirements: formal founding (subset of FOPL) extending RDF(S) by new constructors (cardinality, etc.) three levels of expressivity:

23. three species of OWL OWL-Full : most expressive, but intractable… all of OWL; plus fully `upward’ compatible with RDF --> any legal RDF/S document is also a legal OWL document OWL-DL : limited to a Description Logic ( SHIQ; fragment of FOPL) --> any legal OWL document is a legal RDF/S document (but not vv ) OWL-Light no enumerated classes, disjointness, and full cardinality

24. W3C only provides specifications: no tools thus far, the OWL/RDF development tools come from academia Tools are very important hiding an awful syntax supporting information management graphical interfaces… Three most used tools Protégé with OWL `plug-in’ http:// protege . stanford . edu /download.html Triple-20 (Prolog based, fast classifier): http://swi.psy.uva.nl/tools OILed http://oiled.man.ac. uk /download. shtml Tools for developing ontologies in OWL

25. …Protégé

26. Reasoning tools The KR tools produce OWL structures. To be able to reason one needs `inference engines’ (plug-ins; built-in) eg FACT, RACER,… Classifier (subsumption): automatic classification of new classes (also: multiple classes) automatic verification of individuals: basis for consistency checking small stuff: inheritance, exclusion, … Special reasoning: implications of: part-of structures and aggregation positions and areas of space and time, etc

27. Use of ontologies (1) Knowledge systems: specifying the ingredients of a knowledge base (CommonKADS) part of a(n articulate) knowledge systems (model based reasoning) (high demands on inference) Information retrieval, knowledge management, …SemWeb information = data * knowledge (DB-schemas vs knowledge models) implied (key) terms in search: sub/super class terms Note: more positives (many more fals ones!) annotating and indexing documents handling large quantities of documents and other sources of information question answering (high demands on interpretation and inference) eg web-bots (lowest price of good stuff) eg legal case description --> which law violated

28. Use of ontologies (2) Semantic basis for dialogue, transaction & translation ontology as common semantic reference (vs data-model) eg alternative to EDI ontology as source for mutual understanding role of common sense in human discourse problem: common sense (upper) ontology (CYC?) role of tacit knowledge in professional/specialized communication core ontologies NB: this is not really new: see information science and professional terminology standards ontology as `interlingua’ for NL-translation (cf Euro-Wordnet)

29. Some relevant examples EPISTLE ( http://www.epistle.ws/ ) European process industry terms for components and processes over 20 years experience about 20 permanent staff number of ISO standards etc. Process Specification Language (PSL) http://www.mel.nist.gov/ psl National Institute of Standards Ontology of `process’ fully axiomatized in KIF (FOPL) The CIDOC Conceptual Reference Model (CIDOC CRM) http://cidoc.ics.forth.gr ISO standard for Cultural heritage terminology UMLS , OpenGALEN, etc DL based medical terminology (see Ch 13, Handbook of DL) For more: see Ontoweb portal (SIG-1), deliverables 3.1 and 3.2

30. developing upper ontologies more than 2500 years in philosophy (eg Aristotle) (see Sowa. `99) IEEE-SUO (Standard Upper Ontology) basic reference ontology for Semantic Web… strong committee and open web/email based communication proposals, workshops huge clash of views, alternatives, discussion major trend: the longer the discussion the larger the disagreement proposal to vote! technical/formal merge of several proposals (SUMO) SUMO: physical and mathematical worlds (Sowa, EPISTLE, …) All upper/foundational ontologies are: a source of disagreement necessary to structure and facilitate core/domain ontologies all established core ontologies have upper ontologies. These upper ontologies have been the source of some major overhauls!

31. Some morals: core ontologies Developing core ontologies is highly successful if: there is a well managed , dedicated, professional organization that is well recognized , but works in small teams rather than by open discussion is able to establish some common upper ontology, ie use abstractions about the field of concern NB: the upper ontology emerges as a side effect! However: only the team members learn from sharing

32. Some morals: upper ontologies Developing upper ontologies appears never to be successful by itself increasing divergence cf philosophy (metaphysics) cf huge mailing list of SUO … but they are necessary! emerging from core ontology development (long time effect) reuse! parts etc. some start….

33. Some morals: sharing vs shared in constructing abstract ontologies I have learned a lot in participating in … SUO Ontoweb, SIG-1 community of legal ontologies so that I have my own ideas of what an upper ontology should look like… and more in particular: what a legal core ontology should look like

34. stop that is for another session…

35. constructing LRI-Core need for fixing recurring concepts in law (mostly: common sense) --> LRI-core has a strong `foundational flavour’ view: corresponding with our common sense intuitions about the physical, mental and social world naïve physics vs qualitative physics `revisionary views’ in philosophy needed: `evidence’ from psychological research cognitive (development) psychology evolutionary psychology

36. Principles from this view Common sense in an evolutionary view starting with animal `understanding’ and action primacy of physical world adapting to environmental `domain specific inference engines’ (deficiencies) physical world: (re-)acting to physical change objects: relatively static keep identity independent of position (-> motion) processes: kinds of changes of objects our knowledge of processes is dependent on sensors/perception what changes occur 1) more frequent and 2) more `speedy’

37. some further principles humans vs/and other animals (mammals) intentional stance consciousness natural language: manipulation of symbols representing metaphorization, `reification’ (beliefs, etc.) these all enable the development of worlds beyond the physical world mental world as a metaphor of physical world distinction between behaviour and planned/desired behaviour roles creating abstract world (`form’) by metaphorizing `instincts’ about the physical world (eg: grasping entities of the same kind, counting, …)

38. …and a very basic principle… the knowledge entities (concepts) are `eternal’ once acquired…(ontogenesis; phylogenesis) they work like Plato’s idols the more abstract, the more `eternal’ (eg circle, point, line..) therefore, there are no `temporary’ concepts in an ontology no `occurent’ (perdurant)/ `continous (endurant) distinction IN the ontology e.g. plans, roles, processes, etc may use time/space as a `resource’, but they exist independent of their instantiation e..g. distinction between plan and plan-execution, norm and action, role and role-taker/performance individuals have life cycles (identity criteria) instances `occur’ in time and space

39. ..however… we need terms to refer to occurrences events and states situations and histories foreground/background, system/environment causation: the glue between events on the canvas of space and time positions, areas, instances, duration time’s arrow

40. five `worlds’ of concepts physical world matter/energy --> object and process mental world metaphor intentional stance communication roles physical and social roles social organization abstract occurence

41. physical world basic `natural’ concepts: energy & matter basic defined concepts: physical object & process both contain mixtures of energy & matter objects are in states (see further: `occurences’) processes are/cause changes (and the source of `causation’) transfer (changing places) changing value transformation (changing type) types of processes mechanics: movement & support are core (cf senses & muscles) thermo-dynamics: heat exchange chemistry: mixing/changing substances biology: life, breathing, growing, illness, …

42. process and object energy matter process object heat electricity force state substance transfer quantity form position??? aggregation transformation change-of-value is-a change-of-substance mass change is-a is-a is-a part-of heat exchange radiation movement

43. …processes in OWL-S….

44. the mental world (1) metaphor of the physical world mappings: energy --> emotion|motivation matter/substance --> thought/content (information) object ---> mental-object (concept,…) container ----> mind, memory process ---> mental-process (thinking, memorizing, …) process --> action transfer ---> speaking exchange ---> communication nb: reasoning-structures vs ontology of reasoning terms (hypothesis, evidence, etc) mind/body `problem’ person has mind; mind is container of mental entities action: will as `force’ NB: this naïve view is incorrect! (but still the accepted wisdom in phil.)

45. the mental world (2): intentional stance intention in philosophy: Husserl --> phenomenology (--> Dennet, Searle, etc) instantiated goal state motivated planned intentional/teleological/functional stance vs causal view physical events are explained by processes (causation) agent initiated events are explained by actions actions have intentions world `predicted by’ plans (concatenations of actions and processes) abductive reasoning: from effect to causal determinants (initial states) vs from initial state to (all possible/predictable) states artifacts (physical): taking a intentional view communication….

46. roles distinguishing between role and role taker cutting - knife (for physical objects/artifacts: often: `function’) student - person roles define complementary relations (property constraints) speaker-hearer, student - teacher these `complementary relations’ explain duty/rights relations in legal theories roles ARE behavioural pre-scriptions requirements for role taking (cf man taking `mother role’) norms, procedures role performance may be assessed against role

47. social roles social roles from teleological view on community behaviour `division and distribution of labour’ knowledge about society: role divisions between constructing roles as artifacts and evolving complexity of social organization social organization as `assemblies’ of roles NB the fact that role-taking has a temporary character does not mean that roles are `occurrences’! confounding role with role taker confounding role with role performance in law: norms addressed to roles; responsibility to role taker; norm violation to role performance

48. where it all happens: the world of occurrences “ And in order to understand how common sense works, there is nothing better than imagining “stories” in which people behave according to its dictates.” (Ecco, 99) (semi-)Platonic view: ideas/concepts make up our understanding of what happens in the real world: understanding as constructing a model of a situation episodic vs semantic memory (psychology) Individuals vs Classes (A-Box/T-Box distinction) time and space as the referential canvas of situations and events

49. the world of occurrences-1 situation 1 structural (topological) descriptions of objects in space

50. the world of occurrences-2 situation 2 inferred: time between situation1 and situation2

51. the world of occurrences-3 events & states of objects desk floor teapot ball T-2 T-1 move/fall move/fall move/fall move/fall break collide move/fall

52. the world of occurrences-4 identifying processes desk floor ball T-2 teapot T-1 move/fall move/fall move/fall move/fall break collide move/fall support support

53. the world of occurrences-5 identifying causation desk floor ball teapot move/fall move/fall move/fall move/fall break collide move/fall support support

54. the world of occurrences-6 limiting causal effects… desk floor ball teapot Why does the desk not move? move/fall move/fall move/fall move/fall break collide move/fall support support

55. summary identifying events by recognizing changes, which are viewed as instances of processes (-types) (cf causal-models, Pearl, 2000) identifying causation (= causal relations between events) identifying states as ongoing processes what happens to the forces (heat, energy,…) that are the resources of processes (mental, qualitative simulation) (cf Michotte, 196x)

56. mapping processes to events something event/ state causation subject role object process (type) force resource role subject role resource role causality space time contextualization instanciation

57. What is the Problem? Consider a typical web page: Markup consists of: rendering information (e.g., font size and colour) Hyper-links to related content Semantic content is accessible to humans but not (easily) to computers…

58. What information can we see… WWW2002 The eleventh international world wide web conference Sheraton waikiki hotel Honolulu, hawaii, USA 7-11 may 2002 1 location 5 days learn interact Registered participants coming from australia, canada, chile denmark, france, germany, ghana, hong kong, india, ireland, italy, japan, malta, new zealand, the netherlands, norway, singapore, switzerland, the united kingdom, the united states, vietnam, zaire Register now On the 7 th May Honolulu will provide the backdrop of the eleventh international world wide web conference. This prestigious event … Speakers confirmed Tim Berners-Lee Tim is the well known inventor of the Web, … Ian Foster Ian is the pioneer of the Grid, the next generation internet …

59. What information can a machine see…            …   …  …

60. Solution: XML markup with “meaningful” tags? <name>   </name> <location>   </location> <date>  </date> <slogan>  </slogan> <participants>   </participants> <introduction>    …  </introduction> <speaker>  </speaker> <bio>  </bio > …

61. But What About… <conf>   </conf> <place>   </place> <date>  </date> <slogan>  </slogan> <participants>   </participants> <introduction>    …  </introduction> <speaker>  </speaker> <bio>  …

62. Machine sees… <  >   </  > <  >   </  > <  >  </  > <  >  </  > <  >   </  > <  >    …  </  > <  >  </  > <  >  </  > <  >  </  > <  >  </  >

63. Need to Add “Semantics” External agreement on meaning of annotations E.g., Dublin Core Agree on the meaning of a set of annotation tags Problems with this approach Inflexible Limited number of things can be expressed Use Ontologies to specify meaning of annotations Ontologies provide a vocabulary of terms New terms can be formed by combining existing ones Meaning ( semantics ) of such terms is formally specified Can also specify relationships between terms in multiple ontologies

64. use cases (OWL/W3C) Web portal information management for interest communities by ontologies (eg Ontoweb: http://www.ontoweb.org/ ) Multimedia collections semantic annotations for collections of images, audio, or other non-textual objects. Corporate web site management Design documentation Agents and services Ubiquitous computing