Textual entailment - Wikipedia

Textual entailment (TE) in natural language processing is a directional relation between text fragments. The relation holds whenever the truth of one text ... Textualentailment FromWikipedia,thefreeencyclopedia Jumptonavigation Jumptosearch Textualentailment(TE)innaturallanguageprocessingisadirectionalrelationbetweentextfragments.Therelationholdswheneverthetruthofonetextfragmentfollowsfromanothertext.IntheTEframework,theentailingandentailedtextsaretermedtext(t)andhypothesis(h),respectively.Textualentailmentisnotthesameaspurelogicalentailment –ithasamorerelaxeddefinition:"tentailsh"(t⇒h)if,typically,ahumanreadingtwouldinferthathismostlikelytrue.[1](Alternatively:t⇒hifandonlyif,typically,ahumanreadingtwouldbejustifiedininferringthepropositionexpressedbyhfromthepropositionexpressedbyt.[2])Therelationisdirectionalbecauseevenif"tentailsh",thereverse"hentailst"ismuchlesscertain.[3][4] Determiningwhetherthisrelationshipholdsisaninformaltask,onewhichsometimesoverlapswiththeformaltasksofformalsemantics(satisfyingastrictconditionwillusuallyimplysatisfactionofalessstrictconditioned);additionally,textualentailmentpartiallysubsumeswordentailment. Contents 1Examples 2Ambiguityofnaturallanguage 3Approaches 4Applications 5Seealso 6References 7Externallinks Examples[edit] Textualentailmentcanbeillustratedwithexamplesofthreedifferentrelations:[5] AnexampleofapositiveTE(textentailshypothesis)is: text:Ifyouhelptheneedy,Godwillrewardyou. hypothesis:Givingmoneytoapoormanhasgoodconsequences. AnexampleofanegativeTE(textcontradictshypothesis)is: text:Ifyouhelptheneedy,Godwillrewardyou. hypothesis:Givingmoneytoapoormanhasnoconsequences. Anexampleofanon-TE(textdoesnotentailnorcontradict)is: text:Ifyouhelptheneedy,Godwillrewardyou. hypothesis:Givingmoneytoapoormanwillmakeyouabetterperson. Ambiguityofnaturallanguage[edit] Acharacteristicofnaturallanguageisthattherearemanydifferentwaystostatewhatonewantstosay:severalmeaningscanbecontainedinasingletextandthatthesamemeaningcanbeexpressedbydifferenttexts.Thisvariabilityofsemanticexpressioncanbeseenasthedualproblemoflanguageambiguity.Together,theyresultinamany-to-manymappingbetweenlanguageexpressionsandmeanings.Thetaskofparaphrasinginvolvesrecognizingwhentwotextshavethesamemeaningandcreatingasimilarorshortertextthatconveysalmostthesameinformation.Textualentailmentissimilar[6]butweakenstherelationshiptobeunidirectional.Mathematicalsolutionstoestablishtextualentailmentcanbebasedonthedirectionalpropertyofthisrelation,bymakingacomparisonbetweensomedirectionalsimilaritiesofthetextsinvolved.[4] Approaches[edit] Textualentailmentmeasuresnaturallanguageunderstandingasitasksforasemanticinterpretationofthetext,andduetoitsgeneralityremainsanactiveareaofresearch.Manyapproachesandrefinementsofapproacheshavebeenconsidered,suchaswordembedding,logicalmodels,graphicalmodels,rulesystems,contextualfocusing,andmachinelearning.[6]Practicalorlarge-scalesolutionsavoidthesecomplexmethodsandinsteaduseonlysurfacesyntaxorlexicalrelationships,butarecorrespondinglylessaccurate.[3]However,evenstate-of-the-artsystemsarestillfarfromhumanperformance;astudyfoundhumanstobeinagreementonthedataset95.25%ofthetime,[7]whilealgorithmsfrom2016hadnotyetachieved90%.[8] Applications[edit] Manynaturallanguageprocessingapplications,likequestionanswering,informationextraction,summarization,multi-documentsummarization,andevaluationofmachinetranslationsystems,needtorecognizethataparticulartargetmeaningcanbeinferredfromdifferenttextvariants.Typicallyentailmentisusedaspartofalargersystem,forexampleinapredictionsystemtofilterouttrivialorobviouspredictions.[9] Seealso[edit] Inferenceengine Semanticreasoner Fuzzylogic References[edit] ^IdoDagan,OrenGlickmanandBernardoMagnini.ThePASCALRecognisingTextualEntailmentChallenge,p.2in:Quiñonero-Candela,J.;Dagan,I.;Magnini,B.;d'Alché-Buc,F.(Eds.)MachineLearningChallenges.LectureNotesinComputerScience,Vol.3944,pp.177–190,Springer,2006. ^Korman,DanielZ.;Mack,Eric;Jett,Jacob;Renear,AllenH.(2018-03-09)."Definingtextualentailment".JournaloftheAssociationforInformationScienceandTechnology.69(6):763–772.doi:10.1002/asi.24007.ISSN 2330-1635.S2CID 46920779. ^abDagan,I.andO.Glickman.'Probabilistictextualentailment:Genericappliedmodelingoflanguagevariability'in:PASCALWorkshoponLearningMethodsforTextUnderstandingandMining(2004)Grenoble. ^abTătar,D.e.a.TextualEntailmentasaDirectionalRelation ^TextualEntailmentPortalontheAssociationforComputationalLinguisticswiki ^abAndroutsopoulos,Ion;Malakasiotis,Prodromos(2010)."ASurveyofParaphrasingandTextualEntailmentMethods"(PDF).JournalofArtificialIntelligenceResearch.38:135–187.arXiv:0912.3747.doi:10.1613/jair.2985.S2CID 9234833.Retrieved13February2017. ^Bos,Johan;Markert,Katja(1January2005)."RecognisingTextualEntailmentwithLogicalInference"(PDF).ProceedingsoftheConferenceonHumanLanguageTechnologyandEmpiricalMethodsinNaturalLanguageProcessing:628–635.doi:10.3115/1220575.1220654.S2CID 10202504.Archivedfromtheoriginal(PDF)on13February2017.Retrieved13February2017. ^Zhao,Kai;Huang,Liang;Ma,Mingbo(4January2017)."TextualEntailmentwithStructuredAttentionsandComposition".arXiv:1701.01126[cs.CL]. ^Shani,Ayelett(25October2013)."HowDr.KiraRadinskyUsedAlgorithmstoPredictRiotsinEgypt".Haaretz.Retrieved13February2017. Externallinks[edit] TextualEntailmentResourcePool vteNaturallanguageprocessingGeneralterms AI-complete Bag-of-words n-gram Bigram Trigram Computationallinguistics Natural-languageunderstanding Stopwords Textprocessing Textanalysis Collocationextraction Conceptmining Coreferenceresolution Deeplinguisticprocessing Distantreading Informationextraction Named-entityrecognition Ontologylearning Parsing Part-of-speechtagging Semanticrolelabeling Semanticsimilarity Sentimentanalysis Terminologyextraction Textmining Textualentailment Truecasing Word-sensedisambiguation Word-senseinduction Textsegmentation Compound-termprocessing Lemmatisation Lexicalanalysis Textchunking Stemming Sentencesegmentation Wordsegmentation Automaticsummarization Multi-documentsummarization Sentenceextraction Textsimplification Machinetranslation Computer-assisted Example-based Rule-based Statistical Transfer-based Neural Distributionalsemanticsmodels BERT Document-termmatrix Explicitsemanticanalysis fastText GloVe Latentsemanticanalysis Wordembedding Word2vec Languageresources,datasetsandcorporaTypesandstandards Corpuslinguistics Lexicalresource LinguisticLinkedOpenData Machine-readabledictionary Paralleltext PropBank Semanticnetwork SimpleKnowledgeOrganizationSystem Speechcorpus Textcorpus Thesaurus(informationretrieval) Treebank UniversalDependencies Data BabelNet BankofEnglish DBpedia FrameNet GoogleNgramViewer ThoughtTreasure UBY WordNet Automaticidentificationanddatacapture Speechrecognition Speechsegmentation Speechsynthesis Naturallanguagegeneration Opticalcharacterrecognition Topicmodel Documentclassification LatentDirichletallocation Pachinkoallocation Computer-assistedreviewing Automatedessayscoring Concordancer Grammarchecker Predictivetext Spellchecker Syntaxguessing Naturallanguageuserinterface Chatbot Interactivefiction Questionanswering Virtualassistant Voiceuserinterface Othersoftware NaturalLanguageToolkit spaCy Retrievedfrom"https://en.wikipedia.org/w/index.php?title=Textual_entailment&oldid=1071100872" Categories:TasksofnaturallanguageprocessingLogicalconsequence Navigationmenu Personaltools NotloggedinTalkContributionsCreateaccountLogin Namespaces ArticleTalk English Views ReadEditViewhistory More Search Navigation MainpageContentsCurrenteventsRandomarticleAboutWikipediaContactusDonate Contribute HelpLearntoeditCommunityportalRecentchangesUploadfile Tools WhatlinkshereRelatedchangesUploadfileSpecialpagesPermanentlinkPageinformationCitethispageWikidataitem Print/export DownloadasPDFPrintableversion Languages العربيةDeutschעברית中文 Editlinks