Statistical Morphological Disambiguation for Agglutinative Languages

DMCA

Statistical Morphological Disambiguation for Agglutinative Languages (2000)

Cached

Download Links

Other Repositories/Bibliography

DBLP

by Dilek Z. Hakkani-Tür , Dilek Z. Hakkani-tfir , Gökhan Tür , Kemal Oflazer , Gskhan Tiir

Venue:	In Proceedings of the 18th International Conference on Computational Linguistics
Citations:	49 - 5 self

BibTeX

@INPROCEEDINGS{Hakkani-Tür00statisticalmorphological,
    author = {Dilek Z. Hakkani-Tür and Dilek Z. Hakkani-tfir and Gökhan Tür and Kemal Oflazer and Gskhan Tiir},
    title = {Statistical Morphological Disambiguation for Agglutinative Languages},
    booktitle = {In Proceedings of the 18th International Conference on Computational Linguistics},
    year = {2000}
}

OpenURL

Abstract

In this paper, we present statistical models for morphological disambiguation in Turkish. Turkish presents an interesting problem for statistical models since the potential tag set, size is very large because of the productive dcrivafional morphology. Xh5 propose to handle this by breaking up the morphosyntactic tags into i nfiectional groups, each of which contains the infiectional that;ufos for each (intermediate) derived form. Our statistical models score the probabiliW of each morhosynl;act, ic tag by considering statistics over the individual inflection groups in a trigram model. Among the three models that we have developed and tested, the simplest model ignoring l.he local mort)hotm:ties within words 1)erfibrins the })est. Our })eat l,rigram model 1)eribrms with 93.95% accurlmy on our test (laLa getting all the 111ol:hosyll[,/(:tic alld selll;Hll;ie features CO1Te(;k. If we are just interested in syntact,i(:ally relevanl, timtin'cs and ignore a very snm.ll set, of semantic features, then l.hc accuracy increases to 95.07%.

Keyphrases

statistical morphological disambiguation statistical model agglutinative language ie feature co1te turkish present hc accuracy increase local mort interesting problem individual inflection group trigram model productive dcrivafional morphology ll set nfiectional group ic tag morphosyntactic tag potential tag morphological disambiguation semantic feature tic alld selll

DMCA