Agreement Methodology and Technology for Social Semantics

Transcription

1 Agreement Methodology and Technology for Social Semantics Robert Meersman with bits and pieces provided by Christophe Debruyne VUB STARLab {meersman June 4-5, 2010 COST Action 0801 on AT, Heraklion, Crete

2 Syntax and Semantics

3 Perhaps, towards a real semantic web? Berners-Lee Semantic Web Cake, 1999 Technology (push) driven, architecture Bottom up from syntax Ontologies as key (runtime) resources Top-down adoption difficult Adoption is application (pull) driven Top down bottom up, with domain facts mined from social process workflow Hybrid ontology drives all layers supports operation at enterprise runtime

4 Definition of Ontology (in Computer Science) A conceptualization is a mathematical construct that contains abstract references to (1) objects, (2) relations, (3) functions, and (4) events as may be observed in a given real world. An ontology is a shared, [first order] logical, computer-stored, specification of such an agreed explicit conceptualization. [Tarski 1908, Gruber 1993, Studer 2000, et al.].

5 Synthetically: Semantics = Agreed Meaning Links symbols in autonomously developed systems to shared reality Agreed among humans as cognitive agents increasingly supported by / replaced by sensors Stored in ontologies key technology for interoperability ontologies data models...but provide annotation for them support both human- and system-based reasoning

6 Make agreement easy, separate concerns: a Lexon roles terms < γ, t1, r1, r2, t2 > Context identifier (points to given resource) < γ, bank, situated_at, of, GPS-location > true in a plausible world, i.e., we can imagine an application that uses this fact type; assume evidence for this in context resource collect these (as type level facts) rapidly and easily; applications commit formally to them later Store, e.g., as 2 linked RDF triples + context URI

7 How, semantics of a (single) Information System?

8 Interoperating Enterprises Employees Salesmen Customers! Nothing new unless needs to communicate with tissue of 3 webs

9 Tri-sortal Network of 3 Networks of Actors: Humans Eyeball Web: data, retrieval Systems Semantic Web++: knowledge, reasoning Enterprises Groups, Communities: why we agree

10 Tri-Sortal Web Semantic Web CSCW Workflows Business Web

11 Interoperation Integration, Federation Autonomous nature of actors to be respected Need/wish to communicate, collaborate Need agreements / contracts / meaningful exchange of concepts

12

13 Bi-Sortal Communities permeating the Web Humans Systems goal-oriented Enterprises SECI (From I. Nonaka & H. Takeuchi 1995) P. De Leenheer, PhD Thesis 2009

14 A social experiment in ontology What is the color of this dot?

15 Dual perspectives needed Human perspective: high level reasoning about shared concepts put humans in the loop natural language contexts System perspective : vocabulary agreements, lexons large volume data access low level reasoning

16 Methodology? why not learn from databases Technology matures involve non-geek experts NL discourse analysis (NIAM, ORM) as used for databases: labels vs. things, etc. Use legacy data / output reports / interviews, abstraction into fact types Lift data models into ontologies, remove applicationspecific context

17 Lifting a Data Model fname lname born cy

18 Methodology (DOGMA) Developing Ontology-Grounded Methods and Applications Communities of users / domain experts own the ontology, not programmers. Make use of discourse, social process and legacy resources Ontologies as approximations of perceived reality at type level! (Guarino & Giaretta 1995) Ontologies evolve: lim Ω(t) = Ω real t Users / domain experts rule at every step Lexons in context

19 Double Articulation : Ontological Commitments in DOGMA Lexon Base Commitment Layer Applications

20 Ontological Commitments in DOGMA Commitment = < Selection, Encoding, Constraints > where Selection = set of lexons with various Context-ids Encoding = reference mapping: Application symbols Lexon terms Constraints = set of Ω-RIDL * statements (expressed in lexon terms) * (D. Trog & R. Meersman 2007)

21

22 VUB STARLab spinoff Collibra NV Founded September 2008 by four STARs felix stijn damien pieter Venture capitalized Redeveloped DOGMA tool suite into SAAS Targeting legacy environments, supply chain management, e- government interoperability issues, Technology provider in STARLab research projects

23 Semantic Web 2.0 Technology Adoption Failure Usability vs. Reusability trade-off Leverage and Generativity (cfr. Zittrain 2008) Role of ontology as part of standardisation process Scalability - technological - marketwise - conceptual / social Innovation vs. sustainable technology adoption Linked Open Data: return to syntax... Semantic Web 1.1? but billions of triples do not necess.ly add up to one iota of semantics Lack of a managed Virtuous Circle between providers and adopters Lack of enterprise-workflow oriented methodology

24 (Ongoing work with Michael Brodie, 2010)

25 Semantic Web Social Web Unexpected (except for Tim BL s 1999 comment ) Unsemantic, as yet (Technologically) un-innovative, as yet Socially innovative Marketwise serendipitous

26 Communities First-class citizens in Social Web Purpose-driven in enterprise context - McGrath, FRISCO, Fundamental when agreement > logical proof Co-evolve with their domain-defining ontologies

27 The power of communities = wisdom of the crowds? (with thanks to Jim Hendler) World s best multilingual spell-checker

28 Next: Hybrid ontologies (R. Meersman & T. Dillon, Proc. IEEE DEST 2010, to appear) Human perspective: high level reasoning about shared concepts put & keep humans in the loop natural language and multimedia contexts and glosses System perspective : vocabulary agreements, lexons large volume data access low level reasoning How to make these coexist usefully in same ontology?

29 Hybrid Ontology Base in DOGMA Ω= <Contexts, References, Lexons> Contexts: lexical resources e.g. texts, glosses lightly pre-analyzed for type-level references document human high-level reasoning References: bridge lexical entity types to lexon terms Lexon Base: type-level basis for formal commitments Tool support: hybrid semantic wikis, ( GOSPL, Debruyne, Reul, Meersman: Proceedings IEEE ITNG Conf. 2010)

30 Formal Model NL Description Viewing - Both formal and informal description co-exist

31 Provenance - Tracking changes to the shared agreement on this knowledge

32 references Name: US_State Nickname: US_State Seq_nr: US_State Location_on_map: US_State (Month_3letter_code, Day_number): Date_in_year Medical_condition_Name: Debilitating_Medical_Condition Product_Name: Cannabis_Product Period_in_days: Period Facility_Name: State-monitored_Dispensary Drug_Name: Cannabis_Product Drug_Name: Drug Drug_Name: Crop Weight_in_oz: Product_Quantity Weight_in_grams: Product_Quantity Person_Name: Governor Person_name: Patient Person_Name: Advocate Person_Group_Name: Patients-Rights_Group Person_Group_Name: Lawmakers

33 (Digital) ecosystem issues Rob Whelan s DEST 10 keynote systems compete/evolve as organisms? nutrients? agents, enzymes? systems: embedded in enterprises enterprises increasingly depend on, and so express DNA of their information systems content: semantics! humans play agent role: language, agreement information systems enterprises co-evolve

34 Summary: Research Vectors for Hybrid Ontologies Revisit discourse analysis, pragmatics, semiotics Model communities as 1st class citizens (IFIP s FRISCO, Stamper s MEASUR, McGrath s groups, de Moor s RENESYS, ) Formalize methodologies based on NL involvement of domain experts (Nonaka, CommonKADS, ) Study enterprise workflows in bi- & tri-sortal communities Upgrading role of legacy systems in enterprises Scalable semantic re-exploitation of RDF and LOD resources Multidisciplinarity and Virtuous Circles of developers and adopters

35

36 Thank you