Trang 1
Fall 2019 Entity Resolution in the Web of Data Vassilis Christophides christop@csd.uoc.gr http://www.csd.uoc.gr/~hy562 University of Crete, Fall 2019 Fall 2019 Describing and Linking Entities: Entity-Centric Applications & Knowledge Bases •1
Trang 3
“Entities” is What a Large Part ofFall 2019 Our Knowledge is About Bilbao photos Bilbao map One single Entity Pane can answer many user queries and satisfy users’ diverse information needs Weather in Bilbao Time in Bilbao Bilbao Population Popular schools in Bilbao Bilbao places to go 4 Push/Pull Techniques for Web ContentFall 2019 Keyword Recommendations search Personalized search User Intent Augmented Domain-aware search matching Semantic search https://www.youtube.com/watch?v=L9CPOltkcJA 5 •3
Trang 4
Core Entities Fall 2019 Organizations Locations Movies Persons 6 What Is A Knowledge Base (KB)?Fall 2019 Comprehensive, machine-readable dbpedia:Stanley_Kubrick descriptions of real-world entities are dbo:birt dbpedia:Manhattan hosted in knowledge bases (KB) hPlace Entity names, types, attributes, – rdf:type foaf:Person relationships, provenance info rdf:type yago:AmericanFilm Directors Entities are described as instances of rdf:type yago:Amateur one or several conceptual types and ChessPlayers relationshipsmay be linked through relationships dbpedia: – The core Semantic Web data model A_Clockwork_Orange_(film) dbo:director dbpedia:Stanley _Kubrick dbo:Work/ “136”facts runtime foaf:name “A Clockwork Orange” 7 •4
Trang 5
Knowledge Bases: Scope Fall 2019 Domain-specific Knowledge Base – Focus is on a well-defined domain: • IMDB for movies, Music-Brainz for music, GeoNames for geo, CIA World Factbook for demographics, etc. Global KBs Domain-specific KBs Global Knowledge Base – Covers a variety of knowledge across domains • Concept-intensive (Intensional): Cyc, WordNet • Entity-intensive (Extensional): DBPedia, Yago, Freebase, Knowledge Graph, Satori Bing, Knowledge Vault 8 Knowledge Curation & Knowledge Fusion: Challenges, Models, and Applications X. L. Dong D. Srivastava SIGMOD'15 Tutorial8 Knowledge Bases: History Fall 2019 Cyc WordNet from humans from algorithms for humans for machines guitarist {player,musician} artist Wikipedia algebraist mathematician scientist • x: human(x) • ( y: mother(x,y) Knowledge Graph • z: father(x,z)) 4.5 Mio. English articles • x,u,w: (mother(x,u) 20 Mio. contributors • mother(x,w) • u=w) 1985 1990 2000 2005 2010 Adapted from Knowledge Bases in the Age of Big Data 9 Analytics F. Suchanek, Weikum VLDB 2014 Tutorial •5
Trang 6
Freebase Fall 2019 46.3M entities 1.5K entity types 2.67B triples 4.5K properties John Giannandrea, Freebase – A Rosetta Stone for Entities 11 CIKM 2011 industry talk The Long Tail of Knowledge Fall 2019 Freebase is large, but still very incomplete: Relation % unknown in Freebase Profession 68% Place of birth 71% Nationality 75% Education 91% Spouse 92% Parents 94% We need automatic KB construction methods Kevin Murphy From Big Data to Big Knowledge CIKM 2013 12 industry talk cikm2013.org/slides/kevin.pdf •6
Trang 7
Google’s Knowledge Graph (GKG)Fall 2019 600M nodes (entities) 1.5K entity types 20B edges (triples) 35K properties The Knowledge Graph is used by Google to enhance its search engine's search results with semantic-search information – GKG is rooted in public KBs such as Freebase, Wikipedia and the CIA World Factbook – It’s also augmented at a much larger scale focusing on a comprehensive breadth and depth of entity descriptions 13 Google Knowledge Vault (GKV)Fall 2019 Probabilistic Knowledge Fusion Evaluate trustworthiness of sources by the correctness of its factual information A. Bordes, E. Gabrilovich Constructing and Mining Web-scale 14 Knowledge Graph KDD 2014 Tutorial •7
Trang 8
Entity-Centric Applications Fall 2019 Combine knowledge regarding an entity from multiple sources to build a rich user experience ! studio Entity-Centric Processing spouse director spouse birth place death place Entity death descriptions place Resolved entities Stanley Kubrick spouse Entity-centric Entity-centric lived-in recommendation search entity entity relationships facts 16 Citizen Journalism: Occupations ofFall 2019 Persons from Panama Papers https://finnaarupnielsen.wordpress.com/tag/wikidata/ •8
Trang 9
Entity-Centric Information Fall 2019 Processing Automated construction of entity descriptions – Information extraction: extract new entities from web/text – Link prediction: add relationships among entities Entity integration and resolution – Knowledge base integration: instance & ontology mappings – Entity resolution: merging or splitting similar entities Entity-centric access interfaces – Augmented search: interpret the meaning of queries using entities and compute answers based on a knowledge base – Entity-based matching: recommend new entities given an entity, a user or a query – Entity-centric summarization: of textual posts in social media 18 Fall 2019 Describing and Linking Entities: Linked Knowledge Bases & The Web of Data •9
Trang 10
The Web of Data Fall 2019 Conceptual Representation Typed Links Typed Links Typed Links entity entity entity entity URIs URIs represent URIs represent URIs represent URIs represent URIs RDFa XML HTML Spreadsheets Statistical Triples SemiStructured Web of Data A Web of things in the world (aka entities), described by data on the Web adapted from Chris Bizer, Richard Cyganiak, Tom Heath, http://linkeddata.org/guides-and-tutorials 20 From Open to Linked Data Fall 2019 21 •10
Trang 11
The Linked Data Principles Fall 2019 Anyone can publish data on the Web regarding real-world entities by respecting a minimal set of syntactic conventions – Use URIs as names for things – Use HTTP URIs so that people and machines can look up those names – When someone looks up a URI, provide a useful description – Include links to other URIs, so that they can discover more things Data becomes self-describing – Applications encountering data described by an unfamiliar vocabulary, they can resolve its URIs and understand the vocabulary terms by their RDFS or OWL definitions 22 Fall 2019 Linking Entities in KBs lmdb:director/8476 dbpedia:Stanley_Kubrick dbpedia:Manhattan lmdb:director_na “Stanley dbo:birt dbpedia:Manhattan rdfs:label “Manhattan” me Kubrick” hPlace rdf:type schema.org rdf:type foaf:Person rdf:type foaf:Person :Place foaf:made lmdb:film/1894 rdf:type yago:AmericanFilm dbo:popula 1,626,159 Directors tionTotal foaf:made lmdb:film/2014 rdf:type yago:Amateur foaf:made lmdb:film/2685 ChessPlayers dbpedia: A_Clockwork_Orange_(film) fbase:m.05ldxl dbo:director dbpedia:Stanley fbase:film.directedBy lmdb:director/8476 _Kubrick fbase:film_cut/runtime “136” dbo:Work/ “136” runtime foaf:name “A Clockwork Orange” foaf:name “A Clockwork Orange” 24 •11
Trang 12
From Data Silos to the Web of DataFall 2019 Thanks to W3C standards we can support: – A uniform and universal access to information – Easy linking and re-usage of information in different contexts – Network effects in adding value to information Data Interoperability at the Web Scale – Connecting data from diverse sources and across domains Openness – Many common things are represented in multiple sources 25 – Discover new sources at run-time by following links The Link Open Data (LOD) CloudFall 2019 1014 Knowledge Bases User-generated 60B Triples Media 649 vocabularies Linguistics Government Publications Cross-domain Geographic Life sciences Social networking Last updated: 2014-08-30 http://lod-cloud.net/ 27 •12
Trang 13
LOD Cloud and Knowledge Bases (KB)Fall 2019 adapted from Suchanek & Weikum tutorial@SIGMOD 2013 29 LOD KBs: Entity Statistics Fall 2019 English version 52 entity types 685 entity types 503K entities 4.58 M entities 6M triples 5.83 M triples 2.795 K properties 123 entity types 1.9 B entities 9.6 B triples 112 properties 1.5 K entity types 46.3 M entities 350 K entity types 2.67 B triples 10M entities 4.5 K properties 120M triples 100 properties adapted from Suchanek & Weikum tutorial@SIGMOD 2013 30 •13
Trang 14
LOD KBs: Entity Interlinking Fall 2019 The LOD cloud diagram is sparse top KBs In- (wrt. in- degree degree) DBpedia 207 Only 56.11% of the KBs link to at least geonames 141 another KB w3.org 117 17.36% of them link to only one other KB (typically DBpedia) 31 Fall 2019 Describing and Linking Entities: Description Quality & Entity Resolution •14
Trang 15
Quality of Entity Descriptions in theFall 2019 Web of Data Given the open and decentralized nature of the Web, reliability and usability of entity descriptions need to be constantly improved – Incompleteness: real world entitles are only partially described in KBs – Redundancy: descriptions of the same real world entities usually overlap in multiple KBs – Inconsistency: real world entities may have conflicting descriptions across KBs – Incorrectness: errors can be propagated from one KB to the other due to manual copying or automated extraction/fusion techniques 35 Various Forms of KBs OverlappingFall 2019 among KBs (inter-overlapping, due to common data sources) not identical descriptions, even if they stem from the same source within the same KB (intra-overlapping, due to data extraction or curation problems) – dbpedia:Robert_Soloway and dbpedia:Spam_King refer to the same individual, Robert Soloway, who was nicknamed the “Spam King” – less often than inter-overlaping 36 •15
Trang 16
Fall 2019 Entity Resolution (ER) The problem of identifying descriptions of the same real-world entity A Clockwork highly similar descriptions Orange Data Variety director Stanley Kubrick somehow similar descriptions 37 From Highly to Somehow SimilarFall 2019 Highly Similar Somehow Similar feature many common tokens in feature significantly fewer common the values of semantically related tokens in attributes that are not attributes always semantically related heavily interlinked sparsely interlinked – mostly using owl:sameAs – using various kinds of predicates predicates usually complementary descriptions usually provide complementary of the same entity in terms of descriptions of the same entity in domains, aspects terms of facets, temporal evolution good for linking – – good for fusing typically met in peripheral KBs typically met in central KBs 38 •16
Trang 17
Fall 2019 Data Variety: Curse or Blessing ? 39 How does ER Improves KB Quality?Fall 2019 KB Completeness: – Linking somehow similar descriptions will increase coverage of entity facts and relationships KB Conciseness: – Merging highly similar descriptions will reduce duplicate entity facts and relationships KB Consistency: – Matching similar descriptions will enable to detect conflicting assertions KB Correctness: – Splitting complex descriptions will facilitate entity repairing 40 •17
Trang 18
Fall 2019 Web-Scale ER: Challenges The two core ER problems, how can we effectively compute entity similarity efficiently resolve sets of entities within or across KBs are challenged by the – important number of KBs (~ hundreds) – large number of entity types & properties (~ thousands) – massive volume of entities (~millions) Large-scale, progressive, multi-type, cross-domain ER –Big Data Volume, Velocity and Variety! 41 Fall 2019 Entity Resolution – Formal Definition Entity resolution: The problem of identifying descriptions of the same entity within or across sources – E = {e1, ..., em} is a set of entity descriptions – M : E ×E → {true, false} is a match function – The resolution of entities in E results in a partition P = {p1, ..., pn} of E, such that: 1.∀ei, ej ∈ E : M(ei, ej) = true,∃pk ∈ P : ei, ej ∈ pk 2.∀pk ∈ P, ∀ei, ej ∈ pk, M(ei, ej) = true each partition contains all the matching only matching descriptions descriptions are in the same partition •18
Trang 19
Entity Resolution - Example Fall 2019 dbpedia:Stanley_Kubrick lmdb:director/8476 dbo:birt dbpedia:Manhattan lmdb:director_na “Stanley hPlace me Kubrick” e3 rdf:type foaf:Person rdf:type foaf:Person e1 rdf:type yago:AmericanFilm foaf:made lmdb:film/1894 Directors foaf:made lmdb:film/2014 rdf:type yago:Amateur ChessPlayers foaf:made lmdb:film/2685 Imdb:co0041067 imbd:location GeoNames:Buckinghamshire imbd:filmography “A Clockwork Orange” e5 dbpedia: A_Clockwork_Orange_(film) fbase:m.05ldxl dbo:director dbpedia:Stanley fbase:film.directedBy lmdb:director/8476 _Kubrick fbase:film_cut/runtime “136” dbo:Work/ “136” runtime foaf:name “A Clockwork Orange” foaf:name “A Clockwork Orange” e2 e4 43 Fall 2019 Matching Graph-Structured Entities dbpedia:A_Cloc linkedMDB:A_Cl e2 kwork_Orange ockwork_Orange e4 director directs dbpedia:Sta freebase:Sta e1 nely_Kubrick nely_Kubrick e3 director director spouse children dbpedia:Kath dbpedia:The freebase:The freebase:Ru arina_Kubrick _Shining _Shining th_Sobotka e7 e5 e6 e8 45 •19
Trang 20
Single (Pairwise) Entity MatchingFall 2019 Based on Content Similarity Matching decisions are independent e1 simc(e1,e3) = Jaccard ( e3 birthPlace Manhattan {Manhattan, Person, director_name “Stanley AmericanFilmDirectors, Kubrick” type Person AmateurChessPlayers}, {Stanley, Kubrick, Person, 1894, 2014, type Person type AmericanFilmDire ctors 2685}) = 0.1 directs film/1894 type Amateur directs film/2014 ChessPlayers thresh = 0.5 directs film/2685 simc: let the content similarity of two descriptions be the Jaccard similarity of their values’ token sets 46 Collective (Joint) Entity MatchingFall 2019 Based on Structural Similarity One matching provides evidence for another dbpedia:A_Cloc linkedMDB:A_Cl e2 kwork_Orange ockwork_Orange e4 director directs dbpedia:Sta freebase:Sta e1 nely_Kubrick nely_Kubrick e3 children director director spouse dbpedia:Kath dbpedia:The freebase:T freebase:Ru arina_Kubrick _Shining he_Shining th_Sobotka e7 e5 e6 e8 47 •20
Hiển thị 20 trang đầu. Tải tài liệu để xem đầy đủ.
Bình luận (0)
Chưa có bình luận nào. Hãy là người đầu tiên!