Capturing emerging relations between schema ontologies on the Web of Data
Download as PPTX, PDF0 likes362 views
This document discusses discovering schema mappings between ontologies on the Web of Data by analyzing instance-level links between classes. The approach constructs two networks of schema mappings based on instances that are co-typed between classes or have declared associations. Analyzing these networks provides insights into how ontologies are used and linked in the Web of Data. The resulting mappings could support tasks like selecting relevant data sources or ontologies for integrating new data.
Capturing emerging relations between schema ontologies on the Web of Data
- 1. Capturing emerging relations between schema ontologies on the Web of DataAndriy NikolovEnrico Motta
- 2. Public linked data“Linking” in the Linked Data cloud:References to instance URIs described in external sourcesSpecial case: identity links between equivalent resourcesLinking Open Data cloud diagram, by Richard Cyganiak and AnjaJentzsch. http://lod-cloud.net/
- 3. MotivationSchema heterogeneity is an obstacle both for creating and for utilising these linksExtracting information on the same topic from different repositoriesDiscovering equivalence links between individualsMotivation for our work: discovering instance-level linksHow to choose the repositories to connect a new one?Which subsets of repositories contain co-referring instances??LinkedMDBTV programsDBPediaFreebasemoviespieces of musicMusicBrainz
- 5. Matching approaches“Top-down”Analyzing schema ontologies and generating alignments (manually or automatically)UMBELUsing CYC as a “backbone”Mapping commonly used schema ontologies“Bottom-up”Inferring schema mappings based on instance-level information
- 6. Our approachConstructing a large-scale network of schema mappingsApplying a light-weight instance-based matcherAnalysing the resulting networkWhat does it tell us about the use of ontologies?
- 7. Motivating factorsPotential use case scenariosDiscovering relevant sources for connectionDiscovering relevant subsets of comparable instancesTolerance to the quality of mappingsA mapping between “strongly overlapping” classes is still useful even if there is no strict equivalence/subsumption
- 8. Instance-based matchingUse of instance-based matchingSome implicit schema-level assumptions cannot be captured using only schema-level evidenceInterpretation mismatchesdbpedia:Actor = professional actor (film or stage)movie:actor = anybody who participated in a movieClass interpretation “as used” vs “as designed”FOAF: foaf:Person = any personDBLP: foaf:Person = computer scientist
- 9. Instance set overlapsCo-typingdbpedia:Artistyago:ItalianComposersDBPediais_ais_adbpedia:Ennio_MorriconeDeclared associationmovie:music_contributordbpedia:Artistmo:MusicArtistis_ais_aLinkedMDBDBPediaMusicBrainz==dbpedia:Ennio_Morriconemovie:music_contributor/2490music:artist/a16…9fdf
- 10. DatasetBillion Triple Challenge 2009about 1.14 billion triplescontainscore LOD repositories (DBPedia, Freebase, Geonames, Musicbrainz, LinkedMDB,…)smaller semantic datasets retrieved by search servers (Falcon-S, Sindice)≈3.6M co-typing-based overlapping pairs of classes≈1M association-based pairs
- 11. Inferring mappingsClassification taskClasses A, B: is there a mapping?Boolean classification type of mappings assigned based on comparing sizes of instance setsFeatures𝑛𝑠1, 𝑛𝑠2: namespaces of class URIs|𝑒𝐴∩𝐵|: size of the overlap𝑒𝐴, 𝑒𝐵: sizes of instance sets|𝑒(𝐴∩𝐵)||𝑒𝐴|,|𝑒(𝐴∩𝐵)||𝑒𝐴|: ratio of the overlapping subset to the complete instance setdirect/indirect: whether classes have instances explicitly declared to be equivalent
- 12. TestTrainingTraining set: 6000 overlapping pairs of classesTest: 10-fold cross-validationTrainingTraining set: 6000 overlapping pairs of classesTest: 10-fold cross-validationApplying2 networks of class mappings
- 13. Observations: class mappingsAssociation-based network: classes involved into the largest number of mappingsHigh-level classes represented concepts covered in many repositories… and describing categories with very fine-grained class decompositionUsually also the most populated onesgeonames:Featurefreebase:people.personyago:PhysicalEntitylinkedmdb:filmumbel:Personakt:Personakt:ArticleReference…“under-linked” ones?
- 14. Observations: class mappingsCo-typing-based network: classes involved into the largest number of mappingsPopular classes reused in many repositories… or in DBPedia… and describing categories with fine-grained class decompositionUsually also the most populated onesfoaf:Personumbel:Persondbpedia:Persondbpedia:FootballPlayerwordnet:Persondbpedia:Albumsioc:WikiArticlegeonames:Feature…
- 15. Links between ontologiesAggregated network: connections between ontologiesMapping-based links between ontologiesAt least 1 mapping between corresponding classes must exist
- 18. Freebase
- 19. UMBEL
- 20. OpenCYC
- 22. Freebase
- 23. UMBEL
- 24. OpenCYC
- 25. DBPediaAssociation-based networkMain factor: topic coveragePopularity for linking is not reflectedData-level: DBPedia has more connections than FreebaseSchema-level: no substantial differenceEffect of exploiting composed links
- 26. Co-typing-based networkMain factor:Popularity for reuseFOAF and WordNet:the most popular DBPedia, YAGO, OpenCYC, UMBELReused for DBPedia instances
- 27. OutcomesPossible usage scenarios for mappingsSelecting suitable sources to connect“LinkedMDB contains more movies than DBPedia – more likely to cover all my instances”Selecting an ontology to reuse to structure new instancesWhich sources use this ontology? Do I want my data to be integrated with them?Other data-driven tasksE.g., exploratory searchGeneric challengesHow to take into account task requirements in ontology matching?Recall vs precision, fuzzy vs exactHow to capture changes in the data?BTC 2009 is almost obsolete by now
- 28. Limitations and future workLimitationsLight-weight matcher can lead to lower quality mappingsOK for our scenario but not othersPre-existing instance-level mappings are not always availableFuture workCombining with schema-based ontology matching techniquesTaking into account properties and complex correspondences
- 29. Questions?Thanks for your attention
- 30. Disjoint but overlappingSpurious owl:sameAs linkdbpedia:Hippocrates(Hippocrates) = bookmashup:9004095748 (Hippocratic Lives and Legends (Studies in Ancient Medicine, Vol 4))Spurious rdf:typeassignmentdbpedia:Celtic_Frost (band) defined as Person in DBPedia (fixed in the current version of DBPedia)Modelling assumptionsdbpedia:Masada describes both the geographical place and the battle