Paper: Lexicalized Hidden Markov Models For Part-Of-Speech Tagging

ACL ID C00-1070
Title Lexicalized Hidden Markov Models For Part-Of-Speech Tagging
Venue International Conference on Computational Linguistics
Session Main Conference
Year 2000
Authors

Since most previous works tbr HMM-1)ased tag- ging consider only part-ofsl)eech intbrmation in contexts, their models (:minor utilize lexical in- forlnatiol~ which is crucial tbr resolving some morphological tmfl)iguity. In this paper we in- troduce mliformly lexicalized HMMs fin: i)art - ofst)eech tagging in 1)oth English and ](ore, an. The lexicalized models use a simplified back-off smoothing technique to overcome data Sl)arse- hess. In experiment;s, lexi(:alized models a(:hieve higher accuracy than non-lexicifliz(~d models and the l)ack-off smoothing metho(l mitigates data sparseness 1)etter (;ban simple smoothing methods.