SKOS: Building taxonomies with minimum ontological commitment
- 1. SKOS: Building taxonomies with minimum ontological commitment Szymon Klarman http://klarman.synthasite.com Grakn Labs Semantic Web London Meetup April 18, 2017
- 2. Simple Knowledge Organisation System • W3C standard (2009)… • …for sharing thesauri, taxonomies, classification schemes and subject heading systems… • …using the Semantic Web data model RDF. https://www.w3.org/TR/skos-reference/
- 3. Simple Knowledge Organisation System Car Motor vehicle Motorcycle "car" @en "automobile" @en "auto" @pl"samochod" @pl skos:broader Car washes Services skos:broaderskos:broader skos:relatedMatch Vehicles skos:inScheme Industry classification skos:inScheme skos:prefLabel skos:prefLabelskos:altLabel skos:altLabel
- 4. Knowledge organisation systems controlled vocabulary Expressiveness taxonomy thesaurus ontologyOntological commitment SKOS OWL/RDFS list of terms hierarchy hierarchy + relations classes instances properties
- 5. Ontological commitment & semantics motor vehicle a wheeled vehicle, powered by an internal combustion engine, that does not operate on rails car a motor vehicle, typically with 4 wheels, able to carry a small number of people
- 6. Ontological commitment & semantics
- 7. Ontological commitment & semantics ?x rdf:type ?y ?y rdfs:subClassOf ?z --------------------------- ?x rdf:type ?z Car Motor vehicle has engine some Combustion engine myTesla rdf:type rdfs:subClassOf rdfs:subClassOf Semantics = logical inference rdf:type rdf:type
- 8. Ontological commitment & semantics Car Motor vehicle myTeslaPhoto subject skos:broader X skos:broader Y = „Y has a broader meaning than X” semantics = shared understanding
- 9. Ontological commitment & semantics Domain: Real-world things Questions: -What are they? -What are they like? Semantics: Set-theoretic OWL/RDFS ontologies Domain: Information resources (and concepts). Question: What are they about? Semantics: No formal semantics. Integrity constraints imposed by the data model specs. SKOS taxonomies RDF data model
- 10. SKOS specification • concepts and labels • semantic relations • SKOS mappings • integrity constraints
- 11. Concepts and labels "car" @en "automobile" @en "auto" @pl"samochód" @pl skos:prefLabel skos:prefLabelskos:altLabel skos:altLabel skos:Concept rdf:type Car "A motor vehicle, typically with 4 wheels, able to carry a small number of people" @en skos:definition
- 12. Semantic relations Car Motor vehicle skos:broader Electric car skos:narrower skos:related Electric vehicle skos:broader skos:narrower
- 13. SKOS mappings Car Motor car skos:exactMatch Vehicles skos:inScheme Road traffic vocabulary Car (automatic) skos:inScheme skos:narrowMatch skos:inScheme skos:broadMatch
- 14. Integrity constraints (entailments) •skos:semanticRelation - domain & range = skos:Concept •skos:related - symmetric •skos:relatedMatch - symmetric •skos:broaderTransitive - transitive, inverseOf skos:narrowerTransitive •skos:broader - inverseOf skos:narrower •skos:broadMatch - inverseOf skos:narrowMatch •skos:narrowerTransitive - transitive, inverseOf skos:broaderTransitive •skos:narrower - inverseOf skos:broader •skos:narrowMatch - inverseOf skos:broadMatch •skos:closeMatch - symmetric •skos:exactMatch - symmetric, transitive
- 15. Integrity constraints (non-entailments) Car Motor vehicle skos:broader skos:broader Vehicle skos:broader Car Car wash skos:related skos:related Cleaning products skos:related
- 16. Integrity constraints (not consistent) "car" @en "automobile" @en skos:prefLabel Car skos:prefLabel "car" @en skos:prefLabel Car skos:altLabel
- 17. Integrity constraints (not consistent) Car Motor vehicle skos:broader skos:related Vehicle skos:broader Car Motor vehicle skos:broader skos:related
- 18. Integrity constraints (not consistent) Car Motor car skos:exactMatch skos:broadMatch Car Motor car skos:exactMatch skos:relatedMatch
- 19. SKOS + OWL/RDFS Car Motor vehicle powered by some Combustion engine myCar rdf:type rdfs:subClassOf rdfs:subClassOf myCarPhoto subject skos:broader skos:Concept rdf:type rdf:type owl:Class rdf:type rdf:type rdf:type
- 20. Curated SKOS taxonomies -AGROVOC Multilingual agricultural thesaurus http://aims.fao.org/vest-registry/vocabularies/agrovoc-multilingual- agricultural-thesaurus - STW Thesaurus for Economics http://zbw.eu/stw/ -UK Public sector vocabularies http://standards.esd.org.uk -EUROVOC European Union multilingual thesaurus http://eurovoc.europa.eu/
- 23. Crowdsourced SKOS taxonomies skos:Concept 1,561,776 skos:broader 3,262,643 skos:related 23,483 skos:exactMatch 307 dct:subject 24,941,650
- 24. Crowdsourced SKOS taxonomies crowdsourced taxonomy + crowdsourced tagging = very noisy data + highly imprecise categorisation But: less ontological committment broader participation knowledge acquisition bottleneck reduced cost of building intelligent applications reduced
- 25. SKOS-based recommendations DBpedia movie recommender: http://52.26.26.74:8890/dbmr
- 26. SKOS-based similarity Horror Movie Comedy skos:broaderskos:broader Monster movie Haunted house movie skos:broaderskos:broader Slapstick skos:broader A B C subject subject subject taxonomy taggedcontent
- 28. Boosting SKOS management Taxonomical knowledge Information resources SKOS taxonomy Discovery •iterative •interactive •semi-automated Pool Party Semantic Suite (Semantic Web Company) DBpedia subjects reconciliation / import Concept extraction Recurrent term suggestion
- 29. Context disambiguation Movies „jaws" @en skos:altLabel „jaws" @en skos:altLabel Anatomy
- 30. Summary SKOS is an RDF-based standard for representing taxonomies, thesauri, etc. •Conceptually simple with minimum ontological commitment: •No complex formal semantics and logical implications •Laverages NL semantics and facilitates sharing it •Lower entry treshold for engaging experts/contributors/users •Structurally simple: •Easy to visualise and navigate •Easy to exploit with graph-based algorithms •Convenient bridge between structured and unstructured data.