240 likes | 386 Views
Towards a Standard for Heterogeneous Ontology Integration and Interoperability. Seoul, South Korea - LaRC, June 2011. Oliver Kutz & Christoph Lange Research Center on Spatial Cognition (SFB/TR 8), University of Bremen, Germany & Jacobs University Bremen, Germany Joint work with
E N D
Towards a Standard for Heterogeneous Ontology Integration and Interoperability Seoul, South Korea - LaRC, June 2011 • Oliver Kutz & Christoph Lange • Research Center on Spatial Cognition (SFB/TR 8), University of Bremen, Germany • & Jacobs University Bremen, Germany • Joint work with • Till Mossakowski (DFKI)- Christian Galinski (Infoterm)
Ontology Interoperability • Critical issues are • Semantic Heterogeneity • Syntactic Heterogeneity • Plurality of structuring & modularity concepts • Plurality of documentation techniques • Plurality of tools, editors, reasoners, etc. • Plurality of (kinds of) services, devices, etc.
Overview • Motivating Examples for the use of the hyperontology framework • Structured Ontology Design • Matchingin networks of ontologies • Relations between ontologies: Refinements, Blending, etc. • Universal logic addresses (onto)-logical pluralism and semantic heterogeneity • Hyperontologies = structured and heterogeneous (networks of) ontologies • A Sketch of a future standard: DOL: Distributed Ontology Language
Structured Ontologies Dolce’s structuringin CASL, showing the import structure, i.e. the modular re-use
Ontology Repositories, e.g. BioPortal, Orate, Colore, Tones: collections of ontologies for different purposes and in various ontology languages. • create new ones out of existing ones by finding synonyms, extracting modules, and merging them together. • Meaning shift and “chinese whispers”. • problem of heterogeneity & scalability • problem of “information overflow” Matching Across Repositories
P(x) Q(x,y) heterogeneous refinement Dolce-LITE Dolce-FOL P(x) Q(x,y) Core R(x,t) ∼ R(x) (forget temporal dimension) definitional extension Projection R(x,t) R(x) U(x,y) Approximation V(z)... S(x,y,z) connection through bridge theory Heterogeneous Refinement of Dolce • Different version of Dolce are available, e.g. in DL and FOL: What is their logical relationship?
Ontological Blending Motivation: Conceptual Blending and metaphor: House + Boat = Houseboat Boat + House = Boatshouse Selectively combining two ontologies whilst preserving common structure (theory).
Pluralism in Ontologies • NCI Thesaurusabout 34.000 concepts arranged in 20 taxonomic trees, reference terminology for cancer research, sub-Boolean description logic EL. • Galenmedical domain ontology, relatively large, but also relatively complex axiomatisation in a more expressive DL, namely OWL-DL. • Dolce, BFO, GUM, GFOFoundational ontologies, first-order, higher-order, first-order modal logic being used. Complex axiomatisations.
Universal Logic Items that can be varied according to universal logic: • Signatures: (non-logical symbols) propositions; predicates; functions, constants, terms. • Grammar: (logical symbols) variables and quantifiers; modalities; identity symbol; substitution. • Models: possible world; domains of discourse; accessibility (counterpart relations) ; object (individual) • Satisfaction: vary the truth conditions for quantifiers; Booleans; Modalities; vary conditions for identity statements, etc. Benefits: Borrowing and combination of logics and reasoners, structuring, etc.
Heterogeneous Ontologies • In order to systematically link and combine ontological modules formulated in different formalisms we need to: • fix a logic graph • give logic translations (institution co-morphisms)
Hyperontology example A hyperontology is a heterogeneously interlinked network of heterogeneous ontology modules. Heterogeneous specification of Mereology
Hyperontologies via Matching O1 O3 • 5 participating ontologies, all connected via matchings. • matching results in a single synset identifying all matched concepts, and inconsistency. • removal of the Graphics ontology can cut synset into 2 distinct ones, can restore consistency. • following more than one orange arrow means playing chinese whispers. O5 O2 O4
The Problem of Module Extraction • JRAO is constructed using fragments of NCI and Galen • NCI, Galen are too large to be imported completely • Import only interesting ‘modules’ • Conservativity:Ensure that the ‘module’ is large enough to cover all relevant information (coverage)Ensure that no new information is added (safety)Add only relevant axioms (minimality)
Workflow & Tool Interoperability yes Ontology Repository consistency check Merged Ontology compute colimit no Falcon User Pellet Hets Alignment Specification match pairwise Modules select matching configuration extract modules Matching Configuration Hyperontology Graph produce formal specification
Hets - The Heterogeneous Tool Set • structured representations (such as V-alignments), reuse/independent development of modules • library of logics/formalisms supported, incl. OWL-DL • various provers connected: incl. for OWL-DL, first-order, higher-order, model checker, etc • computation of colimits • checking for conservativities
DOL - Distributed Ontology Language • general purpose framework for ontology interoperabilitylibrary of logics/formalisms supported, incl. most ontology languages • well-defined formal semanticspairs of languages have common target ontology languageApplication T(O) of translation to ontology part of DOL syntax DIF: XML- and RDF-based interchange formats Mapping two ontology languages into a third
DOL - Distributed Ontology Language • support for various module languages as well as one universal lingua francaexplicit module extractioninternalise ontology mappings (first class citizens)make ontology translations available in the language • distributed ontologies in terms of both • different internet locations and • different ontology languages. Mapping two ontology languages into a third
Embedded Ontology Documentation • … but also for human users of an ontology (make ontologies comprehensible) • Knowledge Engineers and Service Developers – reuse! • End Users – when services expose ontology documentation (“labels” and more) as online help Ontological Structuring and Modularity is not only for machines …
Documentation State of the Art • SKOS (Simple Knowledge Organization System): an OWL ontology with some non-OWL axioms • “documented” in HTML manual, and unstructured source comments
Documentation Features Unsupported so far • Informal subsets of an ontology (not yet explicitly modularized) • Subterms of complex axioms • Literate Programming:natural language and formal expressions freely interwoven ⇒ generate ontology and manual from same source
Documentation in DOL • Use existing annotation facilities where possible • In non-XML ontology languages, can't embed documentation⇒ “special” comments, or external, non-intrusive RDF standoff markupReuse existing documentation vocabularies(e.g. OMV = Ontology Metadata Vocabulary) • How to identify subjects? E.g. “the first three axioms”? How to do that in text-oriented ontology languages? – Use XPointer!
Conclusions • Ontologies are widely used to enable interoperability • Currently no unified framework for ontologyinteroperability. • Apply the state of the art in modularity, structuring and documentation, as developed e.g. in software engineering • Enable synchronisation and orchestration of interoperable services • OntoIOp (Ontology Integration and Interoperability) is being proposed in ISO/TC 37/SC 3 in order to fill this gap.