73
Lexical Semantics, Distributions, Predicate-Argument Structure, and Frame Semantic Parsing 11-711 Algorithms for NLP 5 December 2017 (With thanks to Noah Smith and Lori Levin)

Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

  • Upload
    others

  • View
    13

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

LexicalSemantics,Distributions,Predicate-ArgumentStructure,and

FrameSemanticParsing

11-711AlgorithmsforNLP

5 December2017

(WiththankstoNoahSmith

andLoriLevin)

Page 2: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

11-711CourseContext

• Previoussemanticslecturesdiscussedcomposingmeaningsofpartstoproducethecorrectglobalsentencemeaning– Themailmanbitmydog.

• The“atomicunits”ofmeaninghavecomefromthelexicalentriesforwords

• Themeaningsofwordshavebeenoverlysimplified(asinFOL):atomicobjectsinaset-theoreticmodel

Page 3: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

WordSense

• Instead,abank canholdtheinvestmentsinacustodialaccountintheclient’sname.

• Butasagricultureburgeonsontheeastbank,theriverwillshrinkevenmore.

• Whilesomebanks furnishspermonlytomarriedwomen,othersaremuchlessrestrictive.

• Thebank isnearthecornerofForbesandMurray.

Page 4: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

FourMeaningsof“Bank”

• Synonyms:• bank1 =“financialinstitution”

• bank2 =“slopingmound”

• bank3 =“biologicalrepository”

• bank4 =“buildingwhereabank1 doesitsbusiness”

• Theconnectionsbetweenthesedifferentsenses varyfrompracticallynone(homonymy)torelated(polysemy).– Therelationshipbetweenthesensesbank4 andbank1 iscalledmetonymy.

Page 5: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Antonyms

• White/black,tall/short,skinny/American,…

• Butdifferentdimensionspossible:–White/Blackvs.White/Colorful

– Oftenculturallydetermined

• Partlyinterestingbecauseautomaticmethodshavetroubleseparatingthesefromsynonyms– Samesemanticfield

Page 6: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

HowManySenses?

• Thisisahardquestion,duetovagueness.

Page 7: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Ambiguityvs.Vagueness

• Lexicalambiguity:Mywifehastwokids(childrenorgoats?)

• vs.Vagueness:1sense,butindefinite:horse(mare,colt,filly,stallion,…)vs.kid:– IhavetwohorsesandGeorgehasthree– IhavetwokidsandGeorgehasthree

• Verbstoo:IranlastyearandGeorgedidtoo• vs.Reference:I,here,thedognotconsideredambiguousinthesameway

Page 8: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

HowManySenses?

• Thisisahardquestion,duetovagueness.

• Considerations:– Truthconditions(servemeat/servetime)– Syntacticbehavior(servemeat/serveassenator)– Zeugmatest:• #DoesUnitedservebreakfastandPittsburgh?• ??Shepoacheselephantsandpears.

Page 9: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

RelatedPhenomena

• Homophones(would/wood,two/too/to)–Mary,merry,marryinsomedialects,notothers

• Homographs(bass/bass)

Page 10: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

WordSensesandDictionaries

Page 11: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

WordSensesandDictionaries

Page 12: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Ontologies

• ForNLP,databasesofwordsensesaretypicallyorganizedbylexicalrelationssuchashypernym (IS-A)intoaDAG

• Thishasbeenworkedonforquiteawhile

• Aristotle’sclasses(about330BC)– substance(physicalobjects)– quantity(e.g.,numbers)

– quality(e.g.,beingred)– Others:relation,place,time,position,state,action,affection

Page 13: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

WordsensesinWordNet3.0

Page 14: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Synsets

• (bass6,bass-voice1,basso2)

• (bass1,deep6)(Adjective)

• (chump1,fool2,gull1,mark9,patsy1,

fallguy1,sucker1,softtouch1,mug2)

Page 15: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

“Rough”Synonymy

• JonathanSafranFoer’s EverythingisIlluminated

Page 16: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

NounrelationsinWordNet3.0

Page 17: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran
Page 18: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Isahamburgerfood?

Page 19: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

VerbrelationsinWordNet3.0

• Notnearlyasmuchinformationasnouns

Page 20: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

FramebasedKnowledgeRep.

• Organizerelationsaroundconcepts

• Equivalentto(orweakerthan)FOPC– Imagefromfuturehumanevolution.com

Page 21: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Stillno“real”semantics?

• Semanticprimitives:Kill(x,y)=CAUSE(x,BECOME(NOT(ALIVE(y))))Open(x,y)=CAUSE(x,BECOME(OPEN(y)))

• ConceptualDependency:PTRANS,ATRANS,…ThewaiterbroughtMarythecheckPTRANS(x)�ACTOR(x,Waiter)�(OBJECT(x,Check)

�TO(x,Mary)�ATRANS(y)�ACTOR(y,Waiter)�(OBJECT(y,Check)

�TO(y,Mary)

Page 22: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Wordsimilarity

• Humanlanguagewordsseemtohavereal-valuedsemanticdistance(vs.logicalobjects)

• Twomainapproaches:– Thesaurus-basedmethods• E.g.,WordNet-based

– Distributionalmethods• Distributional“semantics”,vector“semantics”

• Moreempirical,butaffectedbymorethansemanticsimilarity(“wordrelatedness”)

Page 23: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Human-subjectWordAssociations

Stimulus:wall

Numberofdifferentanswers:39

Totalcountofallanswers:98

BRICK160.16STONE90.09PAPER70.07GAME50.05BLANK40.04BRICKS40.04FENCE40.04FLOWER40.04BERLIN30.03CEILING30.03HIGH30.03STREET30.03...

Stimulus:giraffe

Numberofdifferentanswers:26

Totalcountofallanswers:98

NECK330.34

ANIMAL90.09

ZOO90.09

LONG70.07

TALL70.07

SPOTS50.05

LONGNECK40.04

AFRICA30.03

ELEPHANT20.02

HIPPOPOTAMUS20.02

LEGS20.02

...

FromEdinburghWordAssociationThesaurus,http://www.eat.rl.ac.uk/

Page 24: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Thesaurus-basedWordSimilarity

• Simplestapproach:pathlength

Page 25: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Betterapproach:weightedlinks• Usecorpusstatstogetprobabilitiesofnodes

• Refinement:useinfocontentofLCS:2*logP(g.f.)/(logP(hill)+logP(coast))=0.59

Page 26: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

DistributionalWordSimilarity

• Determinesimilarityofwordsbytheirdistribution inacorpus– “Youshallknowawordbythecompanyitkeeps!”(Firth1957)

• E.g.:100kdimension vector,“1”ifwordoccurswithin“2lines”:

• “Whoismyneighbor?”Whichfunctions?

Page 27: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Whoismyneighbor?• Linearwindow?1-500wordswide.Orwholedocument.Removestopwords?

• Usedependency-parserelations?Moreexpensive,butmaybebetterrelatedness.

Page 28: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Weightsvs.justcounting

• Weightthecountsbytheapriorichanceofco-occurrence

• Pointwise MutualInformation(PMI)

• Objectsofdrink:

Page 29: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Distancebetweenvectors

• Comparesparsehigh-dimensionalvectors– Normalizeforvectorlength

• Justusevectorcosine?

• SeveralotherfunctionscomefromIRcommunity

Page 30: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Lotsoffunctionstochoosefrom

Page 31: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Distributionally SimilarWords

31

Rumvodkacognacbrandywhiskyliquordetergentcolaginlemonadecocoachocolatescotchnoodletequilajuice

Writereadspeakpresentreceivecallreleasesignofferknowacceptdecideissueprepareconsiderpublish

Ancientoldmoderntraditionalmedievalhistoricfamousoriginalentiremainindianvarioussingleafricanjapanesegiant

Mathematicsphysicsbiologygeologysociologypsychologyanthropologyastronomyarithmeticgeographytheologyhebreweconomicschemistryscripturebiotechnology

(fromanimplementationofthemethoddescribedinLin.1998.AutomaticRetrievalandClusteringofSimilarWords.COLING-ACL.Trainedonnewswiretext.)

Page 32: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Human-subjectWordAssociations

Stimulus:wall

Numberofdifferentanswers:39

Totalcountofallanswers:98

BRICK160.16STONE90.09PAPER70.07GAME50.05BLANK40.04BRICKS40.04FENCE40.04FLOWER40.04BERLIN30.03CEILING30.03HIGH30.03STREET30.03...

Stimulus:giraffe

Numberofdifferentanswers:26

Totalcountofallanswers:98

NECK330.34

ANIMAL90.09

ZOO90.09

LONG70.07

TALL70.07

SPOTS50.05

LONGNECK40.04

AFRICA30.03

ELEPHANT20.02

HIPPOPOTAMUS20.02

LEGS20.02

...

FromEdinburghWordAssociationThesaurus,http://www.eat.rl.ac.uk/

Page 33: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Recentevents(2013-now)

• RNNs(RecurrentNeuralNetworks)asanotherwaytogetfeaturevectors– Hiddenweightsaccumulatefuzzyinfoonwordsintheneighborhood

– Thesetofhiddenweightsisusedasthevector!

Page 34: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

RNNs

Fromopeni.nlm.nih.gov

Page 35: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Recentevents(2013-now)

• RNNs(RecurrentNeuralNetworks)asanotherwaytogetfeaturevectors– Hiddenweightsaccumulatefuzzyinfoonwordsintheneighborhood

– Thesetofhiddenweightsisusedasthevector!• Compositionbymultiplying(etc.)–Mikolov etal(2103):“king– man+woman=queen”(!?)

– CCGwithvectorsasNPsemantics,matricesasverbsemantics(!?)

Page 36: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

36 SemanticProcessing[2]

SemanticCases/ThematicRoles

• Developedinlate1960’sand1970’s

• Postulatealimitedsetofabstractsemanticrelationshipsbetweenaverb&itsarguments:thematicroles orcaseroles

• Insomesense,partoftheverb’ssemantics

Page 37: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

37 SemanticProcessing[2]

ThematicRoleexample

• Johnbrokethewindowwiththehammer• John:AGENTrolewindow:THEMErolehammer:INSTRUMENTrole

• ExtendLFnotationtousesemanticroles

Page 38: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

38 SemanticProcessing[2]

ThematicRoles

• IsthereaprecisewaytodefinemeaningofAGENT,THEME,etc.?

• Bydefinition:– “TheAGENTisaninstigatoroftheactiondescribedbythesentence.”

• Testingviasentencerewrite:– Johnintentionally brokethewindow– *Thehammerintentionally brokethewindow

Page 39: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

39 SemanticProcessing[2]

ThematicRoles[2]

• THEME– Describestheprimaryobjectundergoingsomechangeorbeingactedupon

– FortransitiveverbX,“whatwasXed?”– Thegrayeaglesawthemouse“Whatwasseen?”(A:themouse)

Page 40: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Breaking,Eating,Opening

• Johnbrokethewindow.• Thewindowbroke.• Johnisalwaysbreakingthings.

• Weatedinner.• Wealreadyate.• Thepieswereeatenupquickly.

• Openup!• Someoneleftthedooropen.• Johnopensthewindowatnight.

Page 41: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Breaking,Eating,Opening

• Johnbrokethewindow.• Thewindowbroke.• Johnisalwaysbreakingthings.

• Weatedinner.• Wealreadyate.• Thepieswereeatenupquickly.

• Openup!• Someoneleftthedooropen.• Johnopensthewindowatnight.

breaker,brokenthing,breakingfrequency?

eater,eatenthing,eatingspeed?

opener,openedthing,openingtime?

Page 42: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

CanWeGeneralize?

• Thematicroles describegeneralpatternsofparticipantsingenericevents.

• Thisgivesusakindofshallow,partialsemanticrepresentation.

• FirstproposedbyPanini,before400BC!

Page 43: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

ThematicRoles

Role Definition ExampleAgent Volitional causer of the event The waiter spilled the soup.

Force Non-volitional causer of the event The wind blew the leaves around.

Experiencer Mary has a headache.Theme Most directly affected participant Mary swallowed the pill.Result End-product of an event We constructed a new building.Content Proposition of a propositional event Mary knows you hate her.Instrument You shot her with a pistol.Beneficiary I made you a reservation.Source Origin of a transferred thing I flew in from Pittsburgh.Goal Destination of a transferred thing Go to hell!

Page 44: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

ThematicGridorCaseFrame

• Example:break– Thechildbrokethevase.<agenttheme>

subjobj

– Thechildbrokethevasewithahammer.<agentthemeinstr >

subjobj PP

– Thehammerbrokethevase.< themeinstr >obj subj

– Thevasebroke.<theme>subj

Page 45: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

ThematicGridorCaseFrame

• Example:break– Thechildbrokethevase.<agenttheme>

subjobj

– Thechildbrokethevasewithahammer.<agentthemeinstr >

subjobj PP

– Thehammerbrokethevase.< themeinstr >obj subj

– Thevasebroke.<theme>subjTheThematicGridorCaseFrameshows

• Howmanyargumentstheverbhas• Whatrolestheargumentshave• Wheretofindeachargument

• Forexample,youcanfindtheagentinthesubjectposition

Page 46: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

DiathesisAlternation:achangeinthenumberofargumentsorthegrammaticalrelationsassociatedwith

eachargument

• Chris gaveabooktoDana. <agentthemegoal>subjobj PP

• AbookwasgiventoDana byChris. <agentthemegoal>PPsubjPP

• ChrisgaveDanaabook. <agentthemegoal>subjobj2obj

• Dana wasgivenabookbyChris. <agentthemegoal>PPobj subj

Page 47: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

TheTroubleWithThematicRoles

• Theyarenotformallydefined.

• Theyareoverlygeneral.

• “agent verb theme withinstrument”and“instrumentverbtheme”...– Thecookopenedthejarwiththenewgadget.

→ Thenewgadgetopenedthejar.

– Susanatetheslicedbananawithafork.→ #Theforkatetheslicedbanana.

Page 48: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

TwoDatasets

• PropositionBank(PropBank):verb-specificthematicroles

• FrameNet:“frame”-specificthematicroles

• Thesearelexiconscontainingcaseframes/thematicgridsforeachverb.

Page 49: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

PropositionBank(PropBank)

• Asetofverb-sense-specific “frames”withinformalEnglishglossesdescribingtheroles

• Conventionsforlabelingoptionalmodifierroles

• PennTreebankislabeledwiththoseverb-sense-specificsemanticroles.

Page 50: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

“Agree”inPropBank

• arg0:agreer

• arg1:proposition

• arg2:otherentityagreeing

• Thegroupagreeditwouldn’tmakeanoffer.

• UsuallyJohn agreeswithMary oneverything.

Page 51: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

“Fall(movedownward)”inPropBank

• arg1:logicalsubject,patient,thingfalling• arg2:extent,amountfallen• arg3:startingpoint• arg4:endingpoint• argM-loc:medium• Sales fellto$251.2million from$278.8million.• Theaveragejunkbondfellby4.2%.• Themeteorfellthroughtheatmosphere,crashingintoCambridge.

Page 52: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

FrameNet

• FrameNetissimilar,butabstractsfromspecificverbs,sothatsemanticframes arefirst-classcitizens.

• Forexample,thereisasingleframecalledchange_position_on_a_scale.

Page 53: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

change_position_on_a_scale

Oil rose in price by 2%It has increased to having them 1 day a month.Microsoft shares fell to 7 5/8.Colon cancer incidence fell by 50% among men.

Manywords,notjustverbs,sharethesameframe:

Verbs:advance,climb,decline,decrease,diminish,dip,double,drop,dwindle,edge,explode,fall,fluctuate,gain,grow,increase,jump,move,mushroom,plummet,reach,rise,rocket,shift,skyrocket,slide,soar,swell,swing,triple,tumbleNouns:decline,decrease,escalation,explosion,fall,fluctuation,gain,growth,hike,increase,rise,shift,tumbleAdverb:increasingly

Page 54: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Conversely,onewordhasmanyframesExample:rise

• Change-position-on-a-scale:OilROSEinpricebytwopercent.• Change-posture:aprotagonist changestheoverallpositionorpostureofabody.

– Source:startingpointofthechangeofposture.

– Charles ROSEfromhisarmchair.

• Get-up:A Protagonist leavestheplacewheretheyhaveslept,their Bed,tobeginorresumedomestic,professional,orotheractivities.GettingupisdistinctfromWakingup,whichisconcernedonlywiththetransitionfromthesleepingstatetoawakefulstate.

– I ROSE frombed,threwonapairofcamouflageshortsanddrovemylittleToyotaCorollatoaconstructionclearingafewmilesaway.

• Motion-directional:Inthisframea Theme movesinacertain Direction whichisoftendeterminedbygravityorothernatural,physicalforces.The Theme isnotnecessarilyaself-mover.

– TheballoonROSEupward.

• Sidereal-appearance: An Astronomical_entity comesintoviewabovethehorizonaspartofaregular,periodicprocessof(apparent)motionoftheAstronomical_entity acrossthesky.Inthecaseofthesun,theappearancebeginstheday.

– Atthetimeofthenewmoon, themoon RISES ataboutthesametimethesunrises,anditsetsataboutthesametimethesunsets.

Eachday thesun's RISE offersusanewday.

Page 55: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

FrameNet

• Framesarenotjustforverbs!• Verbs:advance,climb,decline,decrease,diminish,dip,double,drop,dwindle,edge,explode,fall,fluctuate,gain,grow,increase,jump,move,mushroom,plummet,reach,rise,rocket,shift,skyrocket,slide,soar,swell,swing,triple,tumble

• Nouns:decline,decrease,escalation,explosion,fall,fluctuation,gain,growth,hike,increase,rise,shift,tumble

• Adverb:increasingly

Page 56: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

FrameNet

• Includesinheritanceandcausationrelationshipsamongframes.

• Examplesincluded,butlittlefully-annotatedcorpusdata.

Page 57: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

SemLink

• Itwouldbereallyusefulifthesedifferentresourceswereinterconnectedinausefulway.

• SemLink projectis(was?)tryingtodothat

• UnifiedVerbIndex(UVI)connects– PropBank– VerbNet– FrameNet

–WordNet/OntoNotes

Page 58: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

SemanticRoleLabeling

• Input:sentence• Output:foreachpredicate*,labeledspansidentifyingeachofitsarguments.

• Example:[agentThebatter]hit[patient theball][time yesterday]

• Somewherebetweensyntacticparsingandfull-fledgedcompositionalsemantics.

*Predicatesaresometimesidentifiedintheinput,sometimesnot.

Page 59: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Butwait.Howisthisdifferentfromdependencyparsing?

• Semanticrolelabeling– [agentThebatter]hit[patient theball][time yesterday]

• Dependencyparsing– [subjThebatter]hit[obj theball][mod yesterday]

Page 60: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Butwait.Howisthisdifferentfromdependencyparsing?

• Semanticrolelabeling– [agentThebatter]hit[patient theball][time yesterday]

• Dependencyparsing– [subjThebatter]hit[obj theball][mod yesterday]

1. Thesearenotthesametask.

2. Semanticrolelabelingismuchharder.

Page 61: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Subjectvsagent

• Subjectisagrammaticalrelation• Agentisasemanticrole

• InEnglish,asubjecthastheseproperties– Itcomesbeforetheverb– Ifitisapronoun,itisinnominativecase(inafiniteclause)

• I/he/she/we/theyhittheball.• *Me/him/her/us/themhittheball.

– Iftheverbisinpresenttense,itagreeswiththesubject• She/he/ithitstheball.• I/we/theyhittheball.• *She/he/ithittheball.• *I/we/theyhitstheball.• Ihittheball.• Ihittheballs.

Page 62: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Subjectvsagent

• Inthemosttypicalsentences(forsomedefinitionof“typical”),theagentisthesubject:– Thebatterhittheball.– Chrisopenedthedoor.– Theteachergavebookstothestudents.

• Sometimestheagentisnotthesubject:– Theballwashitbythebatter.– Theballswerehitbythebatter.

• Sometimesthesubjectisnottheagent:– Thedooropened.– Thekeyopenedthedoor.– Thestudentsweregivenbooks.– Booksweregiventothestudents.

Page 63: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

SemanticRoleLabeling

• Input:sentence

• Output:segmentationintoroles,withlabels

• Examplefrombook:• [arg0 TheExaminer]issued[arg1 aspecialedition][argM-tmp yesterday]

Page 64: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

SemanticRoleLabeling:HowItWorks

• First,parse.

• Foreachpredicatewordintheparse:– Foreachnodeintheparse:• Classify thenodewithrespecttothepredicate.

Page 65: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

YetAnotherClassificationProblem!

• Asbefore,therearemanytechniques(e.g.,NaïveBayes)

• Key:whatfeatures?

Page 66: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

FeaturesforSemanticRoleLabeling

• Whatisthepredicate?• Phrasetypeoftheconstituent• Headwordoftheconstituent,itsPOS• Pathintheparsetreefromtheconstituenttothepredicate

• Activeorpassive• Isthephrasebeforeorafterthepredicate?• Subcategorization(≈grammarrule)ofthepredicate

Page 67: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Featureexample

• Examplesentence:[arg0 TheExaminer]issued[arg1 aspecialedition][argM-tmp

yesterday]

• Arg0features:issued,NP,Examiner,NNP,path,active,before,VP->VBDNPPP

Page 68: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Example

Figure20.16:ParsetreeforaPropBank sentence,showingthePropBank argumentlabels.Thedottedlineshowsthepath featureNP ↑ S ↓ VP ↓ VBD forARG0,theNP-SBJconstituentTheSanFranciscoExaminer.

Page 69: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

AdditionalIssues

• Initialfilteringofnon-arguments

• Usingchunkingorpartialparsinginsteadoffullparsing

• Enforcingconsistency(e.g.,non-overlap,onlyonearg0)

• Phrasalverbs,supportverbs/lightverbs– takeanap:verbtakeissyntacticheadofVP,butpredicateisnapping,nottaking

Page 70: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Twodatasets,twosystems

• ExamplefrombookusesPropBank

• Locally-developedsystemSEMAFORworksonSemEval problem,basedonFrameNet

Page 71: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

PropBank vsFrameNet

Page 72: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Shallowapproachestodeepproblems

• Formanyproblems:– Shallowapproachesmucheasiertodevelop• Asin,possibleatall forunlimitedvocabularies

– Notwonderfulperformanceyet• Sometimesclaimedtohelpaparticularsystem,butoftendoesn’tseemtohelp

– Definitionsarenotcrisp• Thereclearlyissomething there,butthegranularityofthedistinctionsveryproblematic

• DeepLearningwillfixeverything?

Page 73: Lexical Semantics, Distributions, Predicate-Argument ...tbergkir/11711fa17/Lexical Semantics and FSP.pdffall guy1, sucker1, soft touch1, mug2) “Rough” Synonymy •Jonathan Safran

Questions?