5 Morphology and Word Formation key concepts Words and morphemes Root, derivational, inflectional morphemes Morphemes, allomorphs, morphs
9 1 Syntagmatic and paradigmatic relations in morphology This book provides an introduction to the field of linguistic morphology It
There is no book that deals adequately with morphology in general linguistic terms and that also takes into account fully up-to-date versions of syntactic and
5 5 Inflection, derivation and the syntax-morphology interface This book provides an introduction to the field of linguistic morphology It
Morphology is the study of word formation, of the structure of words Some observations about words and their structure:
(inflectional morphology) word formation (lexical morphology) Morphology is often referred to as grammar, the set of rules governing words in a language
7 jui 2018 · Morphology Francis Katamba 1 Introduction 1 1 THE EMERGENCE OF MORPHOLOGY Although students of language have always been aware of the
BOOIJ Geert, The Grammar of Words: An Introduction to Morphology (2nd edition), Oxford University e-mail, online, Web page, Website, and download
2 Subdomains of Morphology 3 Properties of Morphemes Morphemes and their shapes Morphological Processes 4 Morphology in Computational Linguistics
Morphology 1 1 How to do morphological analysis (or any other kind of linguistic analysis) Morphology is the study of word formation – how words are
![[PDF] Introduction to Morphology - WordPresscom [PDF] Introduction to Morphology - WordPresscom](https://pdfprof.com/EN_PDFV2/Docs/PDF_7/78791_7morphology.pdf.jpg)
78791_7morphology.pdf uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
IntroductiontoMorphology
LinguisticsforComputerScientists
Session4
AntskeFokkens
DepartmentofComputationalLinguistics
SaarlandUniversity
03October2009
AntskeFokkensMorphology1/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Outline
1IntroductiontoMorphology
Introduction
Whataremorphemes?
2SubdomainsofMorphology
3PropertiesofMorphemes
Morphemesandtheirshapes
MorphologicalProcesses
4MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AntskeFokkensMorphology2/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
Outline
1IntroductiontoMorphology
Introduction
Whataremorphemes?
2SubdomainsofMorphology
3PropertiesofMorphemes
Morphemesandtheirshapes
MorphologicalProcesses
4MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AntskeFokkensMorphology3/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
WhatisMorphology?
Morphologyisthestudyofformandstructure.
Inlinguistics,itgenerallyreferstothestudyofformand structureofwords.
AntskeFokkensMorphology4/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
Whatismorphology?
ThetermMorphologycanrefertothreedifferentthings
aDescriptionofthebehaviourofmorphemesandhowthey arecombined. bDerivational,inflectionalandcompositionalprocessesof wordformationoccurringinaspecificlanguage. e.g."GermanhasarichermorphologythanEnglish" cDescriptionofsuchwordformationprocesses.
AntskeFokkensMorphology5/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
WhatareMorphemes?
Morphemes
Morphemesareminimalmeaning-bearingunits:
e.g.talkedcontainstwomorphemes:talkand-ed(past).
Form-functionpairs(sound/sign-meaning)
Basicunitsofmorphology
Morphemesarethe"buildingstones"ofphrases
AntskeFokkensMorphology6/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
Whystudymorphology?(1/2)
Oneofthemainpropertiesoflanguagearethe
sound/meaningpairs Whenanalyzinglanguage(orlearningaforeignlanguage), wecan'tsimplylistallexpressions:thereisaninfinite numberofthem! Sowecomposeexpressionsintosmallerunits:usuallyinto phrasesandwords(syntax)
AntskeFokkensMorphology7/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
Whystudymorphology?(2/2)
Canweusewordsasbasicsound/meaningunits?
Problems:
1Definitionofwordsisunclear
2Wordscanbecomposedofmanycomponentsthat
contributetomeaningand/orgrammar SeveralapplicationsinComputationalLinguisticsbenefitfrom morphologicalanalysis(morelater)
AntskeFokkensMorphology8/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
WordsandMorphemes
Therearetwomainusagesofthetermword:
1Surfaceform(spokenorwrittenrepresentation)
2Abstractform(lemmaordictionaryentry,
e.g.bareinfinitivesinEnglish,nominativesingleformof nounsinLatin) Theclassofformsrepresentingawordindifferentcontexts iscalledalexeme e.g.sing={sing,sings,sang,sung,singing}
BasedonCrysmann2006
AntskeFokkensMorphology9/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
Adefinitionofwords?
Wordscanbedescribedasunitsoflanguage(either
sequencesofsounds,orsigns)thatfunctionasmeaning bearers.Butthisisafuzzynotion,e.g.: talkedinshetalkedexpressesboth"talking"andpast tense.
Ismoreorlessoneword,oraretherethreewords?
Astructuralistsolution:morphemes
AntskeFokkensMorphology10/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
Alanguage:
11-112phonemes
↓
4,000-10,000morphemes
↓
Aninfinitenumberofsentences
AntskeFokkensMorphology11/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
MorphsandMorphologicalAnalysis
Therealisationsofmorphemesarecalledmorphs:
e.g.Englishpluralmorpheme: [NUMBERpl]:-s,-es,-en,-∅ boy-s,box-es,ox-en,sheep
Thesedifferentrealisationsofthesamemorphemeare
calledallomorphs.
Morphologicalanalysis
Segmentationofexpressionsintobasicunits(mostly
startingfromword-level). Classificationofthesebasicunitsaccordingtofunction.
BasedonCrysmann2006
AntskeFokkensMorphology12/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
Typesofmorphemes
FreeMorphemes
Freemorphemescanoccurindependently.Free
morphemesarecommoninbothEnglishandGerman. e.g.boy,sing
BoundMorphemes
Boundmorphemesmustbeattachedtoanother
morpheme,andcannotbeusedindependently. e.g.[NUMBERpl]-s→boys
BasedonCrysmann2006
AntskeFokkensMorphology13/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
Typesofboundmorphemes
Typicalboundmorphemesare:
affixes(boy+s,talk+ed) clitics(French:jenesaispas,jeandnecannotoccur withoutaverb) roots(Spanishhabl-needsanendingindicatingperson, number,mode,etc.)
AntskeFokkensMorphology14/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
Formatives
Morphemesareform-meaningpairs,butnotallsegmental
formshaveanidentifiablemeaning:
Formativesareformswithoutidentifiablemeaning
e.g.LinkingelementsinGermancompounds:
Geburt+s+tag(Birthday),Schwan+en+hals(swanneck).
BasedonCrysmann2006
AntskeFokkensMorphology15/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Introduction
Whataremorphemes?
PseudoMorphemes
Pseudo-morphemesorcranberrymorphemesare
specialcasesofformatives.
Theyaresegment-ablepartofacomplexword,butdonot
haveanindependentmeaning: e.g. cran+berry,rasp+berry re+ceive,con+ceive
BasedonCrysmann2006
AntskeFokkensMorphology16/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Outline
1IntroductiontoMorphology
Introduction
Whataremorphemes?
2SubdomainsofMorphology
3PropertiesofMorphemes
Morphemesandtheirshapes
MorphologicalProcesses
4MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AntskeFokkensMorphology17/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
AreasofMorphology
Wedistinguish:
Wordforming:
Derivationalmorphology
Compounding
Inflection
AntskeFokkensMorphology18/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
DerivationalMorphology
allowstobuildcomplexwordsbycombiningboundand freemorphemes. Derivationaloperationsareperdefinitionoptional,i.e.not requiredbysyntacticcriteria.
AntskeFokkensMorphology19/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Changesmadebyderivationalmorphemes
(a)semantics, e.g.[clear]→[un+[clear]]=unclear (b)syntacticcategory, e.g.[derive] V →[[[derive] V +ation] N +al] Adj =derivational (c)valencyofaverb, e.g.[qaw]'itbreaks'→[t+[qaw]]'hebreaksit'(Havasupai) (d)severalfromtheabove,e.g.[understand] V → [[understand] V +able]=understandable
AntskeFokkensMorphology20/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Compounding
allowstobuildcomplexwordsbyjuxtapositionoffree morphemes. [[sale]+s+[man]],[[dish]+[washer]].
Productivecompoundingresultsinaninfinitelexicon.
8 < :
English
German
Havasupai
9 = ; 8 < : phonetics phonology morphology 9 = ; 8 < : teacher researcher student 9 = ;
BasedonCrysmann2006
AntskeFokkensMorphology21/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
InflectionalMorphology(1/2)
Inflectionisrequiredbysyntacticcriteria,e.g.anEnglish verbmusthavetense. Itmarksgrammatical(=morpho-syntactic)distinctions:
Conjugation(verbalcategories):
1person,number,gender
2tense,aspect,mood,agreement
Declination(nominalcategories)
case,number,gender,degree,definiteness
BasedonCrysmann2006
AntskeFokkensMorphology22/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
InflectionalMorphology(2/2)
Meaningor,atleast,thegeneralconceptis(generally)not changed,thoughwhen,whoorwhatandsometimes where,howandwhethermaybespecifiedbyinflectional morphemes.
Thereareboundandfreeinflectionalmorphemes:
go[TENSEpast]:went go[TENSEfuture]:willgo
BasedonCrysmann2006
AntskeFokkensMorphology23/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Inflection - paradigm
Inflectionalmorphologyistypicallyorganisedinparadigms.
Paradigm
"Asetofformshavingthesameroot/stem,oneofwhichmust beselectedinacertainsyntacticenvironment"(definition basedon[Crystal(1997)](p.277)and[Payne(1997)](p.26))
AntskeFokkensMorphology24/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Paradigm-anexample
Forinstance,Germanconjugation:
presentNUMBERpastNUMBER singularpluralsingularplural
1.dehn-edehn-en1.dehn-tedehn-te-n
2.dehn-stdehn-t2.dehn-te-stdehn-te-t
3.dehn-tdehn-en3.dehn-tedehn-te-n
TakenfromCrysmann2006
AntskeFokkensMorphology25/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Outline
1IntroductiontoMorphology
Introduction
Whataremorphemes?
2SubdomainsofMorphology
3PropertiesofMorphemes
Morphemesandtheirshapes
MorphologicalProcesses
4MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AntskeFokkensMorphology26/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
SomeBasicNotions
Root:anunanalysableform,expressingthebasiclexical
contentofaword.Alsodefinedas'whatisleftofa complexformwhenallaffixesarestripped'.
Stem:consistsofatleastaroot.
Itcancontain(an)derivationalaffix(es).
Ininflectionalmorphology,stemisgenerallydefinedasthe root+athematicvowel.
Base:aformtowhichanaffixmaybeadded.Abasemay
besimplex(root)orcomplex(root+affixes).
AntskeFokkensMorphology27/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
MorphologicalProcesses
Basescanbealteredbythefollowingprocesses:
Affixation
Prefixation
Suffixation
Circumfixation
Infixation
StemModification
Substitution(vowelmutation,suppletion)
Subtraction
SuprasegmentalModification
Tone
Stress
AntskeFokkensMorphology28/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Affixation
Affixesareboundmorphemes
Theirpositionisfixedwithrespecttothebase
aprefixprecedesthebase im-possible asuffixfollowsthebase want-ed acircumfixsurroundsthebase ge-dehn-t aninfixisplacedwithinthebase f-um-ikas'becomestrong',fikas'bestrong'(Bontok,
Philippines)
Affixationcanbearecursiveprocess
Prefixesandsuffixesaremostfrequent
cross-linguistically
AntskeFokkensMorphology29/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Affixation(cont)
Wordscanhaveaninternalstructure(seenextslide)
Theorderofapplicationcanbesignificant,e.g.
[in-[describe-able]],[[*in-describe]-able] [[un-do]-able]vs[un-[do-able]]
Constraintsonmorphemeorderaredescribedby
morphotactics
Morphotacticscanbedeterminedby
wordsyntax(e.g.indescribable) lexicalstrata non-im-partialvs.in-non-partial
BasedonCrysmann2006
AntskeFokkensMorphology30/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Internalstructureofmotorizability
N motor N\V ize V V\A able A A\N ity N (Sproat(1992),p.84)
AntskeFokkensMorphology31/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Typesofaffixationalprocesses
Affixation
constant string continuous base prefixsuffixcircumfix discontinuous base continuous affix infix discontinuous affix transfix copied string reduplication (Crysmann2006)
AntskeFokkensMorphology32/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Infixation
Aninfixisacontinuousaffixthatattacheswithinthebase
InfixationisrareinEuropeanlanguages
Infixationisoftenmotivatedbyprosodicfactors
Tagalogplacesaffixesinthebasetoavoidclosedsyllables (i.e.syllablesthatendinaconsonant) um-+sulat→sumulat sulat+reduplication:susulatandsumusulat um-+aral→umaral Infixationcanalsobepurelymorphologicallyconditioned: e.g.Udi(Nakh-Daghestanian,Azerbaijan)infixation:
RootTransitiveIntransitive
boxbo-ne-x-saboilsbox-ne-saboils uku-ne-k-saeatsuk-ne-saisedible
BasedonCrysmann2006
AntskeFokkensMorphology33/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Transfixation
Athesegmentofatransfixinterleaveswiththebase's
segment(i.e.bothbaseandaffixarediscontinuous) TransfixationiscommoninSemiticlanguages(e.g.Arabic andHebrew)
Thefollowingformsarederivedfromtherootktbin
Maltese
TransfixWordGloss
-i-e-kiteb'hewrote' -i-ukitbu'theywrote' mi-u-miktub'written' -ie-ktieb'book' -o-akotba'books'
BasedonCrysmann2006
AntskeFokkensMorphology34/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Modification
Morphologicalprocessescaneffectsteminternal
segments
TheGermanvowelmutation("umlaut"and"ablaut")are
typicalexamplesofsuchaprocess
Umlaut:
Phonologicallypredictablesegmentalalternation(e.g. vowelfrontinginGerman) a→ä(Wald,Wälder("forest,forests")) u→ü(Mutter,Mütter,("mother,mothers")) o→ö(tot,Tödlich("dead,deadly"))
Ablaut:
Phonologicallyunpredictablesegmentalalternation
gehen,ging,gegangenvssehen,sah,gesehen
BasedonCrysmann2006
AntskeFokkensMorphology35/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Exampleofasuprasegmentalmorpheme
InSabaot(Nilotic,Kenya&Uganda)usesadvanced
tonguerootandnormalvowelsasmorphemiccontrast.
Thisprocessmaybeappliedtotheentireword,asinthe
examplebelow: (1)k ` ccmnyccncct ´ e ka-a-mnyaan-aa-tε-ATR
PAST-1SG-be.sick-STAT-DIR-IMPERF
"Iwentbeingsick(butIamnotsicknow)" (2)k ´ a ´ amny ´ a ´ an ´ a ´ at´ε ka-a-mnyaan-aa-tε
PAST-1SG-be.sick-STAT-DIR
"Ibecamesickwhilegoingaway(andIamstillsick)" (Payne1997,p.29)
AntskeFokkensMorphology36/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Suppletion
Suppletionrefersto'stemreplacement':averbhasmore
thanonestemwhichareusedindifferentcontexts.
InmanyEuropeanlanguages,suppletionoccurswiththe
verb'tobe',e.g.inEnglish,theverbusesthreehistorically differentroots: am,are,is was,were be (Payne,1997)
AntskeFokkensMorphology37/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
SubtractiveMorphology(1/2)
Subtractivemorphologymeansthatpartofthestemis
omittedtomarkamorphologicalprocess.
ForinstanceKoasati(aMuskogeanlanguage,spokenin
theUS):
SingularPluralGloss
pitaf-fi-npit-li-ntosliceupthemiddle lasap-li-nlas-li-ntolicksomething acokcana:-kalnacokcan-ka-ntoquarrelwithsomeone obakhitip-li-inobakhit-li-ntogobackwards
DatatakenfromSproat(1992)
AntskeFokkensMorphology38/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
SubtractiveMorphology(2/2)
Theshapeofthebasecannotbepredictedfromthe
derivedform
SubtractiveMorphologyisproblematicfortheories
assumingthatmorphologyconsistsoftheadditionof morphemes
BasedonCrysmann2006
AntskeFokkensMorphology39/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Reduplication
Reduplicatedmorphemesareformedbyreduplicating
(partof)thebase.
Intotalreduplicationtheentirebaseiscopied,though
minorchangesmayoccur,e.g.([Kiparsky(1987)](p.
115-117)
Indonesian:
orangorangorang 'man''men'
Javanese:
BaseHabitual-RepetitiveGloss
balibolabali'return' udanudanudεn'rain'
BasedonCrysmann2006
AntskeFokkensMorphology40/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
SuprasegmentalMarking
Stress
Englishverb-nounderivations:
VerbNoun
produceproduce permitpermit importimport insultinsult discountdiscount Tone
Chicheˆwa:
FormTense/aspect
ndi-ná-fótokozasimplepast ndi-na-fótókozarecentpast ndí-nâ:-fótókozaremotepast ndí-ma-fotokózápresenthabitual ndi-ma-fótókozapasthabitual
AntskeFokkensMorphology41/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
MorphophonologicalProcesses(1/2)
Theenvironmentofmorphemescaninfluencetheir
appearance(phonologicaland/orgraphemicalternations)
MorphophonologicalAlternations
Assimilation
Homographicnasalassimilation
iN+possible→impossible iN+complete→incomplete iN+resistable→irresistable
Epenthesis:wish+s→wishes
Graphemicalternations:
y+s∼ies
BasedonCrysmann2006
AntskeFokkensMorphology42/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
MorphophonologicalProcesses(2/2)
Theenvironmentinfluencingthemorpheme'sformneed
notbedirectlyadjacenttothemorpheme Harmonyrulesimposeidentityofsoundfeatures(typically vowelfeatures)
E.g.Finnishvowelharmony
lowmidhigh backvowelsaou frontvowelsäöü neutralvowelsei taivas+ta→taivasta(*taivastä) lyhyt+ta→lyhyttä(*lyhytta)
AntskeFokkensMorphology43/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
(Morpho)phonologicalrules [ChomskyandHalle(1968)]proposephonologicalrulesto derive"surface"morphemesinTheSoundPatternof
English(SPE)
Theywereformalizedas(ordered)context-sensitive
rewriterules: a→b/v_w e.g.iN-→im-/_m
AntskeFokkensMorphology44/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
(Morpho)phonologicalrules
Therewasastrongbelievethatrelatedmorphemesareall
derivedfromthesameunderlyingrepresentation,evenif thisformneveroccursonthesurface(e.g.divineand divinitywouldcomefromtherootdivIn)
Theapproachdidnottakegeneralphoneticconstraints
withinthelanguageinaccount,nordiditaddressrulesand tendenciesinmorphemestructures
AntskeFokkensMorphology45/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Declinationofpuella
Latindeclinationofanounofthefirstdeclination:
caseNUMBER singularplural
NOMpuellapuellae
GENpuellaepuellarum
DATpuellaepuellis
ACCpuellampuellas
ABLpuellapuellis
AntskeFokkensMorphology46/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
Syncretism/exponence
Weobserveboth:
syncretism:thesameformisusedtoexpressdifferent featurecombinations. e.g.inthedeclinationofpuella: -ae:GENorDATsingular,orNOMplural -a:NOMorABLsingular -is:DATorABLplural exponence:therelationbetweenformandfunctionis m:n: multi-exponence(cumulation):oneformexpresses severalfunctions.
Here:-amexpressesbothaccusativeandsingular
Extendedexponence:inge-dehn-t,ge-and-texpress
onefunctiontogether.
AntskeFokkensMorphology47/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
MorphologicalProperties - Synthesis
Synthesis:thenumberofmorphemesthattendtooccurwithin aword.
Inisolatinglanguageswordstendtoconsistofonlyone
morpheme.(e.g.Chineselanguages)
Polysyntheticlanguagesareknownforthelargenumber
ofmorphemesthatmayoccurinasingleword.For instance,theQuechuaandInuitlanguages.Thefollowing exampleisfromYup'ik: (3)tuntussuqatarniksaitengqiggtuq tuntu-ssur-qatar-ni-ksaite-ngqiggte-uq reindeer-hunt-FUT-say-NEG-again-3gg-IND 'Hehadnotyetsaidagainthathewasgoingtohuntreindeer' ([Payne(1997)],p.28)
AntskeFokkensMorphology48/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
MorphologicalProperties - Fusion(1/2)
Fusion:thenumberofmeaningunitsthatarefoundinone
morphologicalshape: Agglutinativelanguageshavelittlefusion:eachmeaning componentisrepresentedbyitsownmorpheme(e.g.
Turkish).
Fusionallanguageshavemorphemesthatexpressmany
meaningunits:e.g.-óinSpanishhablóexpresses indicativemode,3rdperson,singular,pasttenseand perfectaspect.
AntskeFokkensMorphology49/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Morphemesandtheirshapes
MorphologicalProcesses
MorphologicalProperties - Fusion(2/2)
InEnglish,bothexamplesofagglutinativemorphemes,and fusionalonescanbefound: agglutinative:anti+dis+establish+ment+arian+ism fusion:vowelchangeinpluralforming(goose/geese)and strongverbs(sing/sang).
Individualmorphemes(rootandnumber/tense)cannotbe
segmentedinchunks,thereforetheseformsarefusional.
AntskeFokkensMorphology50/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
Outline
1IntroductiontoMorphology
Introduction
Whataremorphemes?
2SubdomainsofMorphology
3PropertiesofMorphemes
Morphemesandtheirshapes
MorphologicalProcesses
4MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AntskeFokkensMorphology51/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
MorphologyinComputationalLinguistics
Morphologyrelatedapplicationsincomputationallinguistics are:
1Analysingcomplexwords,definingtheircomponentparts:
anti+dis+establish+ment+arian+ism
2Analysisofgrammaticalinformation,encodedinwords:
sings sing[PERSON3,NUMBERsingular,TENSEpresent]
AntskeFokkensMorphology52/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
MorphologicalProcessing
Inflection
lemmatisation/stemming extractionofgrammatical(morpho-syntactic)features (preprocessingforparsing) Stateoftheart:finitestatetechnology(tobediscussed)
Reductionoflexiconsize(English2:1,German5:1,
Finnish/Turkish>200:1)(Crysmann2006)
AntskeFokkensMorphology53/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
MorphologicalProcessing(cont)
DerivationalMorphology
Semi-productivityisstillachallenge
Rule-basedapproachestendtosufferfromover-generation
CompoundAnalysis
Importantforlanguageswithproductivecompounding
Additionaltask:bracketing
AntskeFokkensMorphology54/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
Whydoweneedmorphology?
Forlinguistictools,suchasparsers:
significantreductionoflexiconsize
Forstatisticalmethods:
reducesunseendata:inamorphologicallyrichlanguage, manywordswillbefoundineachpossibleform,evenina largetrainingcorpus. Machinetranslationrunsintoproblems,inparticularwhentranslating fromamorphologicallypoortoamorphologicallyrichlanguage.This isexpectedtobecomea'hottopic'inMT
Stateoftheart:FiniteStateTransducers
AntskeFokkensMorphology55/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
Non-deterministicFiniteAutomata(NFA)
Definition
Anon-deterministicfiniteautomatonisaquintuple(Q,Σ,
δ,q
0 ,F),where
Qisafinitesetofstates
Σisafinitesetofsymbols
δisatransitionfunctiondelta:Q×Σ→Q,
suchthatforeachq i ∈Qandeachσ∈Σ,thereisaq j suchthatδ(q i ,σ)=q j ,whereq j isanon-finalsinkstate, unlessσislicitatstateq i q 0 ∈Qisauniqueinitialstate
F⊆Qisasetoffinalstates
Atworse,aNFA'scomplexityisexponentialatwordlength
BasedonCrysmann2006
AntskeFokkensMorphology56/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
eren ε st m r s ε ε
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
eren ε st m r s ε ε
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
eren ε st m r s ε ε
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
eren ε st m r s ε ε
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
eren ε st m r s ε ε
Failure
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
eren ε st m r s ε ε
Backtracking
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
er e n ε st m r s ε ε
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
eren ε st m r s ε ε
Failure
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
eren ε st m r s ε ε
Backtracking
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
er en ε st m r s ε ε
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
er e n ε st m r s ε ε
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AnexampleofaNFA
Germanadjectives
klein+er+es 1234
eren ε st m r s ε ε
Accepted!
BasedonCrysmann2006
AntskeFokkensMorphology57/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
DeterministicFiniteAutomata(DFA)
SowhatabouttheworsecaseexponentialcomplexityofNFA? DeterministicFiniteAutomata(DFA)arelinearatworsecase ForeachNFA,thereisalwaysanequivalentDFA(Hopcroftand
Ullman1979)
DFA,Definition
Adeterministicfiniteautomatonisaquintuple(Q,Σ,δ,q 0 ,F), where
Qisafinitesetofstates
Σisafinitesetofsymbols
δisatransitionfunctionδ:Q×Σ→Q,
q 0 ∈Qisauniqueinitialstate
F⊆Qisasetoffinalstates
AntskeFokkensMorphology58/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
FromNFAtoDFA
ForeachNondeterminsticfinitestatemachine,thereisan equivalentdeterministicfinitestatemachine
Steptotake:
1Expandedgesthattakemorethanoneinputcharacter
2Eliminateε-edges(byaddingalternativeedges)
3Constructpowerautomaton(recursivelycombinestates
reachedbythesameinputsymbol)
AntskeFokkensMorphology59/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
Expandingmultiplesymboledges
q
0start
q 1 q 2 q 3 ε er st e ε s m n r ε q
0start
q 1a q 1b q 1 q 2 q 3 ε e s r t e ε s m n r ε
BasedonCrysmann2006
AntskeFokkensMorphology60/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
Eliminatingε-edges
q
0start
q 1a q 1b q 1 q 2 q 3 ε e s r t e ε s m n r ε q
0start
q 1a q 1b q 1 q 2 q 3 ε e s e r t e ε s m n r ε
AntskeFokkensMorphology61/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
Eliminationofεedges
q
0start
q 1a q 1b q 1 q 2 q 3 e s e r t e ε s m n r ε q
0start
q 1a q 1b q 1 q 2 q 3 e s e r r t t e ε s m n r ε
BasedonCrysmann2006
AntskeFokkensMorphology62/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
Eliminationofεedges
q
0start
q 1a q 1b q 1 q 2 q 3 e s e r r t t e s m n r ε q
0start
q 1a q 1b q 1 q 2 q 3 e s e e r r t t e e s m n r ε
BasedonCrysmann2006
AntskeFokkensMorphology63/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
Constructingapowerautomaton
q
0start
q 1a q 1b q 1 q 2 q 3 e s e e r r t t e e s m n r {q 0 } start {q 1a ,q 2 ,q 3 } {q 1b } {q 1 ,q 3 }{q 2 ,q 3 }{q 3 } e s r m,s,n t e m,s,n,r
BasedonCrysmann2006
AntskeFokkensMorphology64/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
FiniteStateTransducers
FiniteStateTransducersarevariantsofFiniteState
Machinesthatacceptslanguageoversymbolpairs
(a:a,a:c)insteadofsinglesymbols
Conventionally,lefthandsymbolscorrespondtolexicon
input,andright-handsymbolstothesurfacestring The∅canappearbothoninputstringandoutputstring, thesymbol"="(or@)standsforthe'any'symbol
FSTscanbeusedtoimplementphonologicalrules
([Johnson(1972)])
BasedonCrysmann2006
AntskeFokkensMorphology65/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
AFiniteStateTransducer
y+s→ies q
0start
q 1 q 2 q 3 y:i =:= ∅:e ∅:s
BasedonCrysmann2006
AntskeFokkensMorphology66/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
Summary
Morphemesareminimalsign/meaningpairs
Morphologicalanalysisplaysaroleinreductionoflexicon size,unknownwordrecognition,etc
Severalmeaningunitscanbemappedinonemorpheme
(multi-exponence)
Phenomenasuchasreduplication,syncretism,
allomorphism,andmorphophonologicalprocessesmake thatmorphemesarenotnecessarilyeasilyrecognizable
FSMformsthestandard(basic)techniquefor
morphologicalanalysis
AntskeFokkensMorphology67/69
uds-logo
IntroductiontoMorphology
SubdomainsofMorphology
PropertiesofMorphemes
MorphologyinComputationalLinguistics
Automata
FiniteStateTransducers
BibliographyI
Chomsky,NoamandHalle,Morris.1968.TheSoundPatternof
English.NewYork,USA:HarperandRow.
Crysmann,Berthold.2006.FoundationsofLanguageScienceand
Technology:Morphology.
http://www.coli.uni-saarland.de/~hansu/courses/FLST05/schedule.html.
Accessedonthe14thofAugust2008.
Crystal,David.1997.TheCambridgeEncyclopediaofLanguage.
Cambridge,UK:CambridgeUniversityPress.
Johnson,C.Douglas.1972.FormalAspectsofPhonological
Description.TheHague,NL:Mouton.
Kiparsky,Paul.1987.ThePhonologyofReduplication.
Payne,ThomasE.1997.Morphosyntax-aguideforfieldlinguists.
Cambridge,UK:CambridgeUniversityPress.
Sproat,Richard.1992.MorphologyandComputation.Cambridge,
USA.MITPress.
AntskeFokkensMorphology68/69