Ihavealwaysbelievedthatthenatureofputationallinguisticsliesinlinguistics.Traditionallinguistibecalledhumanlinguistics,becauseitsresearchobjectivefocusesonpeopleandunication.Therefore,itisnotlogiputationallinguisticstobepletelydivorcedfromtraditionallinguistics,arettrendinthefield.Oherhand,Iadvocateforthestudyofputationallinguisticresearchtoreturntothefual—studyinghumannguageskillsfirst,ahinkinghowtomaketheseabilitiescalcuble.Specifically,thereareteywork:First,Iwillmakefulluseofthelinguistiowledgeofhumanswehaveachievedsofar,andmakethisinformationputational;Sed,Iwilluseputerstodiscoverbroader,deeperandmoreaccuratelinguistiowledge.Thisinformationisusefulforbothhumanunications,andputersce.Basedontheaboveideas,Iampreparingtocarryoutthefollowingstudiesiure.
1:puterizationofHumanLinguistiowledge
Therearethousandsofnguagesintheworld.Throughouthistory,atleasttensofthousandsofnguageshaveexisted.Infact,thecurrentknowledgebaseofhumanlinguistiditshistoryhasbeesohuge,itisnoteasyforasinglepersontoreview.Thankfully,putershavebettermemoriesthanhumans,sotheyhaveamorepowerfubilitytoholdthislinguistiowledgebase.Therefore,myworkwillstudyhowtotransformsuchhumandiariesintoaputationalformat.Iwillsoexamiicexpressionandtheapplicationofadiary.Thecoreofthenguageissemantidtheessenceofadiaryisasemanticset.Thesemanticexpressionofhumandiaryisoftencaughtinthedilemmaofthedefinitionofadefinitionalwordcycle.Iofacilitatetheuseofadiaryforputers,evenforhumans,futurediariesmustbreakthedefinitioncycle.Tothisend,Iwillestablishasemantichierarchy.Atthebottomlevel,thereareasmalmountofverybasic,simplesemantics.Then,levelbylevel,usingbottom-upmethods,eachbuildsmoreandmoreplexsemantics.Thatmeansthatlowerlevelsemantibeusedtodefineupperlevelsemantics,butupperlevelsemantinotbeusedtodefinelowerlevelorthesamelevelsemantics.Iwillstudytwotralissues.First,Iwilldetermiomofthesemantiddefihembymeansotherthanwords;Sed,Iwilldescribetheinsideproceduresformultiplesimplesemanticstostruplexsemanticsbymeansotherthanwords.
2:ExpltheMeismofChildrenLearningLanguage
Humannguageabilityisnotairelyinbornability.Evenifsomenguageskillsareborn,thesepriseonlyaverysmallpartoftheentirehumannguageability.Therefore,justasthekeytoartificialintelligenceliesinmaelearning,thekeytoexplorehumannguageskillsisnotthenguageitself,buttheabilitytolearnthisability,thatis,toexplorehowhumanbeiousenguage.Itisreizedthattherearetwowaystuage,oneasababylearhernguage,thesedasadultslearnaseguage.Itbeinferredthattheabilitytolearnamueisfarstrohantheabilitytolearnaseguage.But,theprocessoflearningamueismuchmoreplicatedthanthelearningofseguage.First,thereisatimeframeforlearningmue,whichisonlyavaibleatthebeginningoflife.Nobodyknowsforcertainwhenmankindstartlearningamue.Butitisreizedthathumansfinishlearningthebasicabilityofusingamueataboutthreeyearsold.Thatmeans,after3yearsold,apersonusuallyfreelyuandthesurroundiences,andfreelyaheirowehislearningprocessispletelyunscious;thatis,humanspletethelearningofamuewhentheyarenotawareofthelearningitself,andthey’treythingoftheprocessafterwards.Sed,thelearningofmueisoftenapahesimultaneousdevelopmentofachild''smentandbehaviorabilities.Itisdifficulttoseparatethenguagelearningfromthewholesystematiclearning.Andthenguagelearningatesoflifeisnguage-basedlearning—thatis,usinganguage(usuallymue)tolearnanuage.Thelearningofamueistheonlyinallearning.Ipntoresearchthesemanticbasisunderlyingmuelearning.Iwillusemanyapproachestoductthiswork,includiudyofaage,animatioion,andimagination.
3:SemanticVectorSpace
Languageisacolleofsymbols,andallofthewords,phrases,sentences,paragraphs,asstituteasymbolspaputersaregoodatcalcution,andsoifasymbolisrepresentedasanumericalvalue,thenguageproblemisexpressedasamathematicalequationoranexpressiohenguageproblembesolvedusingtheputationalpoweroftheputer.Therearealreadysomematurewaystoexpresswords’formasnumericalvectors,butthestudyofthevectorrepresentationofwordmeaningsandgrammaticalrulesisverylimited.Thecoreofnguageissemantics;allwords,phrases,sentences,paragraphs,asareinteoexpresssemantics.Itisceivablethatallsemanticsofanguagevergetoformasemanticspace.Simirtothemethodofwords,semanticspaalsobeexpressedasavectorspace.Thatis,eachsemanticscorrespondstoanumericalvector.Therearealotofquestionsworthstudying:First,Ihaveamethodtodetermihesemanticspace.Iwilldesignaneffitalgorithmtoproduceeachsemanticvector.Obviously,theiicvectorbesideredfromaword-basedperspective,ahewordbinationsproducephrases,sentences,andernguageuocorrespondtoothersemanticvectors.Iwillstudythealgorithmstomapwordstotheirsemanticvectors.Therearealotofpolysemouswordsinanguage,soawordislikelytomaptomultiplesemanticvectors.Sedly,ifthebinationofwordsisbasedongrammaticalrulesaistraints,thenIwillstudytheoperationsiorspacethatthesegrammaticalrulesaistraintscorrespondto.sideracaseinwhiadjectiveandanounbiogetanounphrasewhosesemanticvectoriscalcutedbyabinatioionontheadjectivesemanticvectorandthesemanticvector.Atthispoint,Iwilldefihisoperationbythesemanticsofadjectiveandnoun.Finally,themappingoflinguisticsymboicvectorsshouldbebidireal.Giveicvector,theremustbetheabilitytostudyhoitbacktothesymbolicspaguage,thatis,tofiofthenguage,whibewords,phrases,sentences,ortext,etc.Ifthisbi-direappingisimplemented,manynaturalnguageprocessing(NLP)taskswillbe.Forexample,futuremaetranstionbeachievedbytwostepmapping.ThetextofnguageAisfirstmappedtothesemanticspadthenmappedfromthesemanticspacetothetextofnguageB.
FutureJobOpportunities
InJanuary2016,IreturnedfromtheUatestoShanghaiandmypositionasanAssociateProfessorintheSchoolofputerSceatFudanUy.IbegantoapplyforfacultypositionsatAmeriiversities.InJuly2016,theStateUyofNewYork(SUNY)atOapresentedmeaformaljobofferforthepositiourerintheDepartmentofMathematics,StatistidputerSce.AlthoughIwasuoacceptthejobofferatthetime,beyondSUa,manyotheruieshaveexpresseddireterestinmyemployment,includingonStateUy,theCityUyofNewYork,IndianaUy,StonyBrookUy,theUyinia,andothers.PleaserefertotheattachedemailsfirmingtheseuniymentorsatWashingtonUyandHarvardandmyotherUScolleaguesarealsoihusiasticsupportofmytinuedreseartheUates.
Baseduponmypriorsuccessinthefieldofputationallinguistidmyoverallreputation,IamcertainofmyabilitytofindsuitableemploymentwithinmyfieldiedStates.Furthermore,myputationallinguisticsresearchisavitallyimportantfieldintheinformatioheextentthatnguageisafunofthehumanmind,aputationaluandingofnguagealsoprovidesinsightintohumanthinkingandintelligence.Singuageisourmostnaturandmostversatilemeansofunication,linguisticallypetentputerswillgreatlyfacilitateouriionwithmaesandsoftwareofallsorts.putationallinguisticsisalsothemosterciallyviablebranchoflinguistics;hundredsofpaheUatesworkonputationallinguistics.sideringtheimpactmyworkhashadthusfar,IamfidentthatmyworkwilltiopositivelyihefieldaedStatesiure.
内容未完，下一页继续阅读