FinnishOTProsody - Fi(n)nish OT

This preview shows page 1 - 13 out of 52 pages.

The preview shows page 11 - 13 out of 52 pages.
Fi(n)nish OT ProsodyFi(n)nish OT ProsodyLauri KarttunenFSMNLP-2005September 1, 2005
OverviewOverviewSuccess of Finite-State MorphologyLexical transducersTwo ways of describing morphological alterationsSequential (Chomsky & Halle 1968)Parallel (Koskenniemi 1983)Optimality Theory (Prince & Smolensky 1993)Yet another two-level modelFinite-state?Lenient compositionFinnish OT ProsodyBasic FactsFinite-state implementation of Kiparsky’s 2003 analysiswith the FST tool (Beesley & Karttunen 2003)ConclusionFinal thoughts
Computational morphologyComputational morphologyAnalysisleavesleaf N Plleave N Plleave V Sg3Generationhang V Pasthangedhung
Two challengesTwo challengesMorphotacticsWords are composed of smaller elements that must be combined in a certain order:piti-less-ness is Englishpiti-ness-less is not EnglishPhonological alternationsThe shape of an element may vary depending on the contextpity is realized as pitiin pitilessnessdie becomes dy in dying
Morphology is regular (=rational)Morphology is regular (=rational)The relation between the surface forms of a language and the corresponding lexical forms can be described as a regular relation.A regular relation consists of ordered pairs of strings.leaf+N+Pl : leaveshang+V+Past : hungAny finite collection of such pairs is a regular relation.Regular relations are closed under operations such as concatenationiterationunion, and composition.Complex regular relations can be derived from simple relations.
Morphology is finite-stateMorphology is finite-stateA regular relation can be defined using the metalanguage of regular expressions.[{talk}|{walk}|{work}][% +Base:0 | %+SgGen3:s|%+Progr:{ing}| %+Past:{ed}];A regular expression can be compiled into a finite-state transducer that implements the relation computationally.
work+3rdSg-->worksk:kt:ta:aa:aw:wo:ol:lr:r+Progr:iε:g+3rdSg:s+Past:eε:dε:n+Base:εGenerationGeneration
talked-->talk+Pastk:kt:ta:aa:aw:wo:ol:lr:r+Progr:iε:g+3rdSg:s+Past:eε:dε:n+Base:εAnalysisAnalysis
Lexical transducerLexical transducerveutvouloir +IndP +SG + P3Finite-statetransducerinflected formcitation forminflection codesvouloir+IndP+SG+P3veutBidirectional: generation or analysisCompact and fastComprehensive systems have been built for over 40 languages:English, German, Dutch, French, Italian, Spanish, Portuguese, Finnish, Russian, Turkish, Japanese, Korean, Basque, Greek, Arabic, Hebrew, Bulgarian, …
How lexical transducers are madeHow lexical transducers are madeLexiconFSTRuleFSTsCompilerfat+Adjr+CompfatteLexical Transducer(a single FST)compositionLexiconRegular ExpressionRulesRegular ExpressionsMorphotacticsAlternations
Two-level rules vs. rewrite rulesTwo-level rules vs. rewrite rulescomposeintersectFSTrule 1rule 2rule n...Surface formLexical formKoskenniemi 1983Intermediate form...Surface formLexical formrule 1rule nrule 1Chomsky&Halle 1968
Rewrite rulesRewrite rulesEpenthesisHarmonyLowering? u: ty? A s? u:tIy ? A s? u: tuy ?as?o:t u y ? a sYawelmaniVowel HarmonyKisseberth 1969

Upload your study docs or become a

Course Hero member to access this document

Upload your study docs or become a

Course Hero member to access this document

End of preview. Want to read all 52 pages?

Upload your study docs or become a

Course Hero member to access this document

Term
Fall
Professor
staff
Tags
Finnish, Constraints, Lenient composition, Finnish OT Prosody, Lexical transducers

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture