程序代写代做代考 flex algorithm junit PowerPoint Presentation

PowerPoint Presentation

LECTURE 13

SemanticRolecLabelling

ArkaitzcZubiaga,c19thcFebruary,c2018

2

 WhatciscSemanticRolecLabelling?

 ThematicRoles:

 FrameNet.

 PropBank.

 ApproaihesctocSemanticRolecLabelling.

 SeleitonalcRestriitons.

LECTUREc13:cCONTENTS

3

 Semanti role labelling (SRL):cforcaccprediiatecorcverbcincac
sentenie,ctaskcofcidentfyingcthematicroles,csuihcascagent,cgoal,c
orcresult.

SEMANTICcROLEcLABELLING

Theme LocationPredicateAgent

4

 SRLciancbecusefulcfor:

 Questoncanswering.

 Maihinectranslaton.

 Informatoncextraiton.

 Tocidentfycsenteniescwithcidentialcmeaning:

 XYZciorporatoncboughtcthecstoik.

 TheycsoldcthecstoikctocXYZciorporaton.

 ThecstoikcwascboughtcbycXYZciorporaton.

SEMANTICcROLEcLABELLING

5

 Givencacsentenie:
PaulcboughtcaciarcfromcSarahcforc£7,000

c

 Identfycprediiate(s)c[i.e.cthecaiton,cgenerallycthecverb]:
PaulcboughtcaciarcfromcSarahcforc£7,000

c

 Identfycthematicroles:

PaulcboughtcaciarcfromcSarahcforc£7,000

SEMANTICcROLEcLABELLING:cEXAMPLE

buyer seller

goods
money

6

 Actypiialcsetcofcthematicroles.

THEMATICcROLES

7

 Most frequently used themati roles:

 AGENT:csubjeitc(aitvecsentenie),cbycXc(passivecsentenie).

 EXPERIENCER:canimatecsubjeitcincaitvecsentenies.

 THEME:cobjeitc(transitvecverb),csubjeitc(non-aitoncverb).

 INSTRUMENT:cwithcX.

 BENEFICIARY:cforcX.

THEMATICcROLES

8

 Problem of themati roles:

 Hard to ireate standard setcofcrolescorcformallycdefnecthem.
c

 Alternatiectocthematicroles:

 PropBank,cfewer,cmorecgeneralisedcsemanticroles.

 FrameNet,cmorecroles,cspeiifictocacgroupcofcprediiates.

THEMATICcROLES

9

 Iniludesc12 domainsc(body,ciogniton,chealth,ctme,…)

 Conduited on frame-by-frame basis:

1.cihoosecsemanticframec(e.g.cCommerie_buy).

2.cdefnecthecframecandcitscframecelementsc(e.g.cBUYER,c
GOODS,cSELLER,cMONEY).

3.clistcprediiatescthatcevokecitc(buy.v,cpurihase.v,cpurihase.n).

4.cextraitcsenteniescforceaihcprediiatecfromcBritshcNatonalc
Corpus.

THEcFRAMENETcPROJECT

10

 Frame Relaton:cdireitedcrelaton between two frames,ci.e.cac
Super_Framec(lesscdependent,cmorecabstrait)candcacSub_Framec
(thecmorecdependent,clesscabstrait).
c

 e.g.cthecframecCommeriial_transaitonchasctwocsub-frames:

 Commerie_goods_transfer

 Commerie_money_transfer

FRAMENET:cFRAMEcRELATIONS

11

 One million word iorpuscannotatedcwithcprediiatecargumentc
struitures.

 DevelopedcatcUniversitycofcPennsilvania.

 Prediiatescareclexiialisedconlycbycierbs.

 Adjunitie arguments:cLoiatve,cTemporal,cManner,cCause,ceti.

PROPOSITIONcBANK:cPROPBANK

12

 Arg0c=cagent

 Arg1c=cdireitcobjeitc/cthemec/cpatent

 Arg2c=cindireitcobjeitc/cbenefaitvec/cinstrumentc/catributec/cendc
state

 Arg3c=cstartcpointc/cbenefaitvec/cinstrumentc/catribute

 Arg4c=cendcpoint

PROPBANK:cARGUMENTcNUMBERS

13

PROPBANK:cADJUNCTS

14

 Theciompanycboughtcacwheel-loadercfromcDresser.

 Arg0:cTheciompany

 Rel:cbought

 Arg1:cacwheel-loader

 Arg2-from:cDresser

PROPBANK:cEXAMPLE

15

 TVcstatonscboughtc“Cosby”crerunscforcreiordcpriies.

 Arg0:cTVcstatons

 Rel:cbought

 Arg1:c“Cosby”creruns

 Arg3-for:creiordcpriies

PROPBANK:cEXAMPLE

16

 BUY SELL PAY

 Arg0:cbuyer Arg0:cseller Arg0:cbuyer

 Arg1:ciommodity Arg1:ciommodity Arg1:cpriiecpaid

 Arg2:cseller Arg2:cbuyer Arg2:cseller

 Arg3:cpriie Arg3:cpriie Arg3:ciommodity

 Arg4:cbenefiiary Arg4:cbenefiiary Arg4:cbenefiiary

PROPBANK:cFLEXIBILITY

17

FRAMENETcVScPROPBANK

18

 Early SRL systemsc(1970s-80s):

 parsercfollowedcbychand-writencrulescforceaihcverb.

 diitonariescwithcverb-speiificiasecframes.
c

 Current systems use maihine learning.

APPROACHEScTOcSEMANTICcROLEcLABELLING

19

AcSIMPLEcMODERNcALGORITHM

20

 Usescparsedctreecascinput.

SEMANTICcROLEcLABELLING:cEXAMPLE

21

1.cPruning:cusecsimplecheuristisctocprunecunlikelycionsttuents.

2. Identiiaton:cacbinarycilassifiatoncofceaihcnodecascanc
argumentctocbeclabelledcorcacNONE.

3.cClassiiiaton:cac1-of-Ncilassifiatoncofcallcthecionsttuentscthatc
wereclabelledcascargumentscbycthecpreviouscstage.

3-STEPcVERSIONcOFcSRLcALGORITHM

22

 Imbalaniecbetween:

 positvecsamplesc(ionsttuentscthatcarecargumentscofc
prediiate).

 negatvecsamplesc(ionsttuentscthatcarecnotcargumentscofc
prediiate).

 Imbalanied data ian be hardcforcmanycilassifers.

 Socwe prune the iery unlikely ionsttuents irst,candcthencusecac
ilassiferctocgetcridcofcthecrest.

WHYcPRUNINGcANDcIDENTIFICATIONcSTEPS?

23

 Algorithm so far ilassiies eierything loiallyc–ceaihcdeiisionc
aboutcacionsttuentciscindependentcofcallcothers.

 But this ian’t be right:cLotscofcglobalcorcjointcinteraitonsc
betweencarguments.

 Acloialcsystemcmayciniorreitlyclabelctwocoverlappingc
ionsttuentscascarguments.

 e.g.clabellingconecionsttuentcARG0cshouldcinireasecthec
probabilitycofcanothercbeingcARG1

FINALcSTAGE:cJOINTcINFERENCE

24

 Reranking:

 ThecfrstcstagecSRLcsystemcproduiescmultplecpossibleclabelsc
forceaihcionsttuent.

 Thecseiondcstagecilassifescthecbestcglobalclabelcforcallc
ionsttuents.

 Ofencacilassifercthatctakescallcthecinputscalongcwithcotherc
featuresc(sequeniescofclabels).

HOWcTOcDOcJOINTcINFERENCE

25

SRL:cNOTcJUSTcENGLISH

26

SEMANTICcROLEcLABELLING:cEXAMPLE

27

 Prunecthecprediiatecandcallcothercelementscthatciniludecit.

SRLcEXAMPLE:cPRUNING

28

 Binarycilassifiatoncofceaihcelement:cargumentcorcnot.

SRLcEXAMPLE:cARGUMENTcIDENTIFICATION

29

SRLcEXAMPLE:cARGUMENTcIDENTIFICATION

30

 Multilasscilassifiatoncofcargumentctype.

SRLcEXAMPLE:cARGUMENTcCLASSIFICATION

31

SRLcEXAMPLE:cARGUMENTcCLASSIFICATION

32

 Featuresctocilassifycarguments:

 Governingcprediiate,ce.g.cissued.

 Phrasectypecthatcitcis,ce.g.cNP.

 Pathcfromcargumentctocprediiate.

 Positon,ci.e.cbeforecorcafercprediiate.

 Voiie,ci.e.caitvecorcpassive.

 Headcwordcofcthecionsttuent.

SEMANTICcROLEcLABELLING:cFEATURES

33

 Contentcword:clexiialisedcfeaturecthatcseleitscancinformatvec
wordcfromcthecionsttuent,cothercthancthechead.

 POSctagcofctheciontentcword.

 POSctagcofcthecheadcword.

 NamedcEnttycilasscofctheciontentcwordc(loiaton,cperson,c
organisaton,…)

 Containscnamedcenttes?cBooleancfeature.

SEMANTICcROLEcLABELLING:cMOREcFEATURES

34

 Firstcwordcofcthecionsttuentc(andcitscPOSctag).

 Lastcwordcofcthecionsttuentc(andcitscPOSctag).

 Lefcandcrightcsiblingcionsttuentclabels.

 Lefcandcrightcsiblingcheadcwordsc(andctheircPOSctags).

SEMANTICcROLEcLABELLING:cMOREcFEATURES

35

 SRL:cAclevelcofcshallowcsemantiscforcrepresentngceventscandc
theircpartiipants.

 Twociommoncarihiteitures,cforcvariousclanguages:

 FrameNet:cframe-speiificroles.

 PropBank:cProto-roles.

 Currentcsystemscextraitcby:

 Parsingcsentenie.

 Findingcprediiates,cilassifyingcassoiiatedcarguments.

SEMANTICcROLEcLABELLING:cSUMMARY

36

 Considercthectwocinterpretatonscof:
Icwantctoceatcsomewherecnearby.

c

1.cIcwantctoceatcincacnearbycplaie.

2.cItciscacplaiecnearbycthecobjeitcthatcIcwantctoceat.
c

 Difiultcforcancalgorithmcifcwecexpeitcthecfollowingcpatern:
eatc+c[food],ce.g.ceatcihiiken

SELECTIONALcRESTRICTIONS

37

 Seleitonalcrestriitonscarecassoiiatedcwithcsenses:

 Thecrestaurantcseriescpizzas.c[servesc+cfood]

 BritshcAirwayscseriescLondon.c[servesc+cplaie]

SELECTIONALcRESTRICTIONS

38

 Commoncpraitiecforcseleitonalcrestriitonscisctocusecthesauric
suihcascWordNet.
c

 e.g.cThe THEMEcofc“eat”cmustcbecWordNetcsynsetc{food,cnutrient}

SELECTIONALcRESTRICTIONS

39

RESOURCES

 FrameNet:
http://framenet.icsi.berkeley.edu/

 PropBank:
http://www.cs.rochester.edu/~gildea/PropBank/Sort/

 NomBank:
http://nlp.cs.nyu.edu/meyers/NomBank.html

http://framenet.icsi.berkeley.edu/
http://www.cs.rochester.edu/~gildea/PropBank/Sort/
http://nlp.cs.nyu.edu/meyers/NomBank.html

40

RESOURCES:cSOFTWARE

 SENNA:
https://ronan.collobert.com/senna/

 SEMAFOR:
http://www.cs.cmu.edu/~ark/SEMAFOR/

 MatePlus:
https://github.com/microth/mateplus

 Open Sesame:
https://github.com/Noahs-ARK/open-sesame

https://ronan.collobert.com/senna/
http://www.cs.cmu.edu/~ark/SEMAFOR/
https://github.com/microth/mateplus
https://github.com/Noahs-ARK/open-sesame

41

ASSOCIATEDcREADING

 Jurafsky, Daniel, and James H. Martin. 2009. Speech and Language
Processing: An Introduction to Natural Language Processing, Speech
Recognition, and Computational Linguistics. 3rd edition. Chapter 22.

Slide 1
Slide 2
Slide 3
Slide 4
Slide 5
Slide 6
Slide 7
Slide 8
Slide 9
Slide 10
Slide 11
Slide 12
Slide 13
Slide 14
Slide 15
Slide 16
Slide 17
Slide 18
Slide 19
Slide 20
Slide 21
Slide 22
Slide 23
Slide 24
Slide 25
Slide 26
Slide 27
Slide 28
Slide 29
Slide 30
Slide 31
Slide 32
Slide 33
Slide 34
Slide 35
Slide 36
Slide 37
Slide 38
Slide 39
Slide 40
Slide 41