Automatic induction of lexical features [Elektronische Ressource] / von Peter Jäger

Description

GutenAutomaticereicInductioneterofundLexicalMainzFPeatures05Inauguraldissertationderzurerg-UnivErlangungondesausAkJ?gerademischhenPhilosophieGradesPhilologieeinesJohannesDr.bphil.,ersit?tvvorgelegtPdemJ?gerFWiesbadenaceterh2011bagReferenUlrict:henProfessorHeidDr.mW04.10.2010alterhBisangTKderoreferen?ndlict:Pr?fung:ProfessoriiDr.forAbstractaThisGrammarthesisvconcernsharticiallyextractsinbtelligenecialtThenaturalfromlanguageestigatedproautomaticallycessingosystemsdescriptionsthatconcisearethecapableunderofylearninghatheonepropofertiesinofidealexicalwitemsable(propofertiesnlikenemenvoferbalthisvaalencylexicalorLearn-inectionaltheclassypmemacquisitionbeenership)builtautonomouslyHead-drivwhilesystems.theyeareoffulllingsystemtheirEnglishtaskshine-readableforwledgewhicahLearn-theylargehajorvaethatbofeenhighlydeploayamedtheinoutthefactsrstdataplace.

Informations

• Publié par: johannes_gutenberg-universitat_mainz
• Publié le: 01 janvier 2011
• Nombre de lectures: 70
• Langue: English
• Poids de l'ouvrage: 4 Mo

Extrait

GutenAutomaticereicInductioneterofundLexicalMainzFPeatures05Inauguraldissertationderzurerg-UnivErlangungondesausAkJ?gerademischhenPhilosophieGradesPhilologieeinesJohannesDr.bphil.,ersit?tvvorgelegtPdemJ?gerFWiesbadenaceterh2011bagReferenUlrict:henProfessorHeidDr.mW04.10.2010alterhBisangTKderoreferen?ndlict:Pr?fung:ProfessoriiDr.G + L + C → S
′G + L + S → L
α
α
thesislearningmentheinprophniqueertiesscratcofacquirelexicaleectivitemsin(propaerties(HPSG)likofesystemvberbalenvthealencyoneorofinectionalvclassofmemthisbwhicership)ofautonomouslyewhilesuctheybaredemandsfulllingsystem,theiredtasksyforcouplewhicacquisitionhAbstracttheydesignhaevcompedevboneenendeploTheyisedexpinisthecorpus.rstlanguageplace.isManwyhal-ofitthesethetasksossiblerequireeciedawithdeepyanalysisconstructionofalanguageainput,t,whictrieshwhethercancriteriabtesubsumedctheharacterizedrule.asesaqualitmappingprototofsucutteranceswhoseintsabgivedenandinputofCstate-of-the-arttoStructureacessingsetySypofvlinguisticallyseriesmotivts,atedthestructureswithwithlargethethehelpmacoftolinguisticknoinformationnew,enconotdedofinfulllsaofgrammaraGbandthatanlexiconofL:descriptionsofhighlycapableenareathatvsystemsamcessingmenprothe(1)conciseTheoutideaofthatfactsunderliesecialindatatelligenthettheselexicalendsacquisitiondecisionsystemsaisoftoformotelligendifylexicalthissystemsscunderhematicparadigm:formLearn-uladesigninThesucdescribhtheaandwyaayypthatforthehsystemsystem,isacquisitionableonentohaexploitetheeeninformationelopencofromdedhinbuiltStoptoonecreatetheaHead-drivnew,PhraseimprGrammaroveprodsystems.vqualitersionofofprotottheelexicon:inlanguageestigatednaturalatoftelligenerimenininarticiallyhconcernssystemthesisfed(2)extractsMoreoavEnglisher,Whiletheideathesisusingclaimshine-readablethatinputaautomaticallysystemlexicalcanwledgeThisnotandwablearedealalargeareora.aothatfourLearn-joronlyisbtoewithconsideredcorpinTtelli-instancegenmatciflengesitconstructingdohessystem,notshouldjustemaktionedea)maximhighumumusageerofpthestructurallecausedarningyoppundersportunitieslexicalintriesC,forbutparserifaiteryisealsobiguitablemanage-totrb)eviseautomaticfalselyofacquiredlexicallexicaltriesknoofwledge.bulkSo,observonelexicalofrequiresthespcentectralofelemenalignmentsc)inreliabilitthisofwenorkdepisonthesystem'sformonulationiiiofα
α
α
ell`enough'resultsinputChapterandselectiond)thegeneralframewpropdetection,ertiesaofaturlanguageamighacquisitiontfourthrendertsomecategoriza-lexicalcomplemenfeaturesationindeterminableonifandtheconceptssystemoftriesthetoANALacquireimplementhemtswithoaprotottoerbalonominalhighandprecision.TheThecon-cornerstoneemenofforthisofdissertationLearn-isosesthenotionsmotivquiredationerandlexicaldevdevelopmenyptd,ofREDUCE,awhicgeneralLearn-theoryhapterofdesignautomaticalexicalexpacquisitiononthate:isofapplicablealencytoinevt/masserypreplanguagetenandamongindepconcludesendenoftandofimproanasypropparticularresearctheoryautomaticofalgrammar.orhaslexicon.expThisallwrelatedorkandisre-dividedforinproptounderstandingvarticialeacquisition.c3hapters.elopsTheprototinedtromethoductorycalledcYZE-LEARN-hapterarstorkcon-htraststsseen.threeThedierenctpresenandthemandutuallyofincompatiblebapproacotstrappingheserimentoconducted(articial)thislexicalypacquisition:lexemecue-basedlearningqueries,vhead-lexicalizedvprobabilistic,contiontexttofreecoungrammarsclasses,andoflearningositionsbsenytialunication.ts,Thenothers.thethesispwithostulationreviewofthetheclusionsLearn-motivitforfurtherdesignvruletsiswpresenasted.osalsThefuturesecondhcthehapterinductionoutlineslexicthefetheoryesthativunderliesbusinessAsuppchaknoforwledgemenytsaIvwishytoconsiderationthanksourceevyperyyonetwhoerconetributedtotowingthisject.thesis.toFirstasoftributedall,withinIIwpatience,ouldtirelesslikmethesistobthankconstruc-mk.ygosupyervisor,forProf.ectDr.thisWbigaltergoBisang,opforunithisproconetintheuouspresensuppthesis.ort,ersonaladvice,thankencouragemenfortderstanding,andWithoutthekup,nadviceumerousbinspir-wifeingoulddiscussions.vMyaccomplished.gratitudetheirextendstivtofeedbacmManythankssecondoutreviewmer,ITProf.colleaguesDr.shoUlricresphandHeid,forforprohisAinthankvoualuableessuggestionstheandenthecommoppyortunittheiryductstovpresenconttomprototyeprotedjecttheatOnthepInstitutnotef?rheartilyMascmhinellefamilySpractheirhun-vanderarbort.eitungtheinbacStuttgart.encouragemenIandalsoofthankytheelootheredmemthisbwersnevofhameyeenthesisvcommittee−α
−α
−α
.duction2:1.1.1.TheLogicLanguage.Engineering69Bottlenec.kA-priori.e.F...Example.72.F...........a..1621.2.Aim.and.StructureNounof.this.Thesis..T.......the...A...biguit...61.a.Learn.....y6.1.31.5.1PreviousalencyResearc.hPon.Computer-aidedALexical.ALearncquisition.....9eature1.3.1.Lexical2.1.2ATcquisition.T.asks......55.wledge.System...1.4.3.Articial.....1.4.4.of........9Wh1.3.2GrammarEvoaluationorkoftsLexical.A.cquisition.Metho.dsLearning.some.......1.erbal.V11.1.3.363DierenVtConstructionStrategies.for.LexicalExampleAescquisitionounds...Theory.Kno........13.1.3.477Cue-basedandStatistical.Approac.hes:.Bren.tF1991ensandesSub-.sequen.t83A.ccoun.ts............1.4.2.Kno.of.Learning.......57.The.of.Lexical.cquisition.......5719Relativ1.3.5DegreeLexicalAmAycquisition.with.Head-lexicalized.Probabilistic.CF.Gs:.Carrol/Ro.oth1.4.51998y.Unication-based.is.Go.d.ramew.for.1..................331.51.3.6bLearningUnication:onIllustrationsthe.Job:.Lexical.A.cquisition.as63aExampleBy-proandductV36V1.3.7andDiscussionoice.......1.5.2.3:.erb.article...........1.5.3.4:.djectiv.and.Comp.......2Inoftro2.1.77.Lexical.wledge..50.1.4.The.Learn.ten.Con...Design.Rule......2.1.1.eatures.F.Structures.............78.Lexical.eatures,.ok.and.yp....55.1.4.1.Exp.erience.andviChallenge−α
−α
−α
−α
max
−α
.h.y.(L.TH)......the86.2.1.4.Symypb.olichemaLanguage.Mo.dels..e.......T...and.........in.in....89the2.1.5.The.Structure127of.Lexical.F.acts..F.......T...2.5.6.2.6.....2.6.2.159...of91.2.1.6.What.it.MeansSetto.Kno124w.a.Lexical125F.eature.V.alue......Barrier..93.2.2NoiseA.dequacy.and131Soundness:.The.P.erspNeutralizationectiv.e135of.the.LanguageNeutralizationEngineer..Consequences.....Observ...A...and...2.6.1.......the...........Lexemes...of...2.7.3.174.....2.4.2.Exp..95.2.2.1.The.OptimalNoiseMo.del........Noise...........2.4.5.............The......95.2.2.2.The129A.dequate.Mo.del......Lexical...........135.V.........Systematic.Neutralization.....2.5.396of2.2.3.Impro.v.emen137ts:LearnT.argeting.the.Optimal.Mo2.5.5delSc.Lexical...in.Space....97Asso2.2.4ositionalitA.Mo.dal.Logiccationsfor.Learn.and.es......cations.ation.....Lexeme.............2.7.1.W98.2.2.5.Grammaticalit.y:2.7.2Whatordthe.System.Kno.wsordAb.out.the.Optimal.Mo.del..123.Noise.the.erience.................2.4.3.in.Grammar..............101.2.2.6.What2.4.4itinMeansLexiconto.A.dmit.a.F.eature.V.alue....127.Eectiv.Noise........102.2.3.T.o.w.ards.a.Logical2.4.6ANoiseccoun.t.of.Learn.yp.T............2.4.7.Scenarios............103.2.3.1.Observ.ations:.Exploiting.Learning2.5OppNeutralizationortunities.......103.2.3.2.Lexical.A.cquisition.Space....2.5.1.eature.alue.................2.5.2.and.otal..106.2.3.3.Induction....136.T.es.Lexical...............2.5.4.for.yp.................145.Generalized108ation2.3.4hemaFRDoAeature.V.alue147RealizationBifurcations.Lexical.cquisition.........148.Lexical.ciations.Comp.y..........1101492.3.5ColloDecisions,.Quan.tication.of.Evidence.and.T.esting.P.o.w.er.1111542.4ColloNoiseand.Observ.Sc.........2.7.Detection.........................159.Detection.Single.ord.............161.Detection123Multiw2.4.1ExpressionsThe.Noise.Mo.del.in.Learn162LexicalMultiweExpressionsHierarcLearn.2.1.3............vii.−α
−α
Seedccoun251t.of.Learn231A1:Probabilistic............In...............237178Requiremen2.8.1forLearning.Do.wn.theduleHierarcouth.y..G.....yp...............the....178.2.8.2.F.eature.V.alue.Characteristics....................Structure.....224..181.2.8.3TheDefaultyF.eatureSeedV.alues..En.....F.....Mo.......3-5:...Language...PET.245.for.3.3.5187.2.8.4.SparseduleData.and.Quan.tizationyEects......de.....Executable........188.2.8.53.5.1ANLE.Algorithm....Seeds.............224.the...........The.y.......225190Seed2.8.6HierarcThe.Statistical.Mo.del3.2.4in.the.Non-ideal.W.orld..Lexical........190.2.8.73.2.6DecisionsVand.Strength.of.Belief..236.ANAL.............Requiremen.dications.G...3.3.2194cessing2.8.8.Cascading243F2-3:oforcus..Requiremen.HoG.....ccoun.........3.4...............The......196.2.8.9.Example..251.e...........252.Options.......3.5...............La..........196.2.9viiiRevision.of.Belief........................3.2.1.of.ER.....................3.2.2200Lexical2.10LaLearnerA.2.8.....Revised..........3.2.3.Lexical.T.e.h.............227.Lexical.Mapping..............206.2.10.1.Reviewing3.2.5theSeedRequirementriests..................233.Default.eature.alues......206.2.10.2.The.Final.V.ersion..3.3.dule.YZE.........................3.3.1.ts.Mo.to.ER..207.3.Analyze,238Learn,NaturalReducePro209Infrastructure3.1.Arc.hitecture.of3.3.3ALRts.Making.Ready.ALR.....3.3.4.t.Making.Ready.ALR.....247.A.ting.Punctuation..............2102483.1.1MoOvLEARNerview..........................3.4.1.La.out....................210.3.1.2.F3.4.2unctionalteractivSpMoecication:.ANAL.YZE..................3.4.3.Program:214.3.1.3.F.unctional.Sp.ecication:.LEARN..256.Mo.REDUCE..................215.3.1.4.F.unctional257SpTheecication:yREDUCE.......................2192573.2Lexical−α
−α
.MoFde..y.311.5.4.Regions.....Questions.The...Maxim.....V.A..............257.3.5.3A.2ExecutableositionsProgram:.Options..P.......A.Learning.....,..........261.4.A.Bo.otstrapping.Exp.erimenPreliminariest.2623294.1.Motiv.ation.andPGoals....296...............Predicativ...310.......wn.....ok.......5.2262.4.2.Design.of.the.Exp.erimen.t..Results.....322.......5.6.......e.o...and.terv...ix....264.4.3.Preliminary4.9Tdicationests..........V.Constructions.........300.V.............304.e.Usage.es..