Abstract. This paper presents a robust method for the construction of collection-specific document models. These document models are variants of the well-known vector space model, which relies on a process of selecting, modifying, and weighting index terms
SyntaxversusSemantics:
AnalysisofEnrichedVectorSpaceModels
BennoStein1andSvenMeyerzuEissen1andMartinPotthast2
Abstract.Thispaperpresentsarobustmethodfortheconstructionofcollection-speci cdocumentmodels.Thesedocumentmodelsarevariantsofthewell-knownvectorspacemodel,whichreliesonapro-cessofselecting,modifying,andweightingindextermswithrespecttoagivendocumentcollection.Weimprovethestepofindextermselectionbyapplyingstatisticalmethodsforconceptidenti cation.Thisapproachisparticularlysuitedforpost-retrievalcategorizationandretrievaltasksinclosedcollections,whichistypicalforintranetsearch.
Wecompareourapproachto“enriched”vector-space-baseddoc-umentmodelsthatemployknowledgeoftheunderlyinglanguageintheformofexternalsemanticconcepts.Primaryobjectiveistoquan-tifytheimpactofapurelysyntacticanalysisincontrasttoasemanticenrichmentintheindexconstructionstep.Asaby-productwepro-videanef cientandlanguage-independentmeansforvectorspacemodelconstruction,whereastheresultingdocumentmodelsperformbetterthanthestandardvectorspacemodel.
Keywordscepts,textcategorization,vectorspacemodel,evaluationconceptmeasures
identi cation,semanticcon-1INTRODUCTION
Eachtextretrievaltaskthatisautomatedbyacomputerreliesonsomekindofdocumentmodel,whichisanabstractionoftheorigi-naldocumentd.Thedocumentmodelmustbetailoredwellwithre-specttotheretrievaltaskinquestion:Itdeterminesthequalityoftheanalysis,and—diametricallyopposed—itscomputationalcomplex-ity.Thoughitsobvioussimplicitythevectorspacemodelhasshowngreatsuccessinmanytextretrievaltasks[11;12;16;15],and,theanalysisofthispaperusesthismodelasitsstartingpoint.
Thestandardvectorspacemodelabstractsadocumentdtowardavectordofweightedindexterms.Eachtermtthatisincludedindderivesfromatermτ∈dbyaf xremoval,whichisnecessarytomapmorphologicalvariantsofτontothesamestemt.Therespec-tivetermweightsindaccountforthedifferentdiscriminativepoweroftheoriginaltermsindandarecomputedaccordingtosomefre-quencyscheme.Themainapplicationofthevectorspacemodelisdocumentsimilaritycomputation.
Inthispaperwefocusontheindexconstructionstepand,inpar-ticular,onindextermselection.Otherconceptsofthevectorspacemodel,suchasthetermweightingschemeoritsdisregardofwordorderareadopted.
ρ.Perhapsthemostinterestingresult
maybeanticipated:Thepositiveeffectofsemanticindextermen-richment,whichhasbeenreportedbysomeauthorsinthepast,couldhardlybeobservedinourcomprehensiveanalysis.
Theremainderofthepaperisorganizedasfollows.Section2presentsataxonomyofindexconstructionmethodsandoutlinescommonlyusedtechnology,andSection3reportsonsimilarityanal-ysisandunsupervisedclassi cationexperiments.
These tools have their own formalisms, semantics ...The delivery of enriched information that results ...Many analysis and design techniques use models to...
2.2 Semantically Enriched Object Information Thinking about mobile smart ... a common understanding of the semantics of the information gathered in ...
generating semantically enriched ontologybased user ...semantics of the concepts and is learned as we ...In Formal Ontology in Conceptual Analysis and ...
generating semantically enriched ontologybased user ...semantics of the concepts and is learned as we ...In Formal Ontology in Conceptual Analysis and ...
热门文档