Word Sense Disambiguation with Distribution Estimation

Yee Seng Chan, Hwee Tou Ng
2005 International Joint Conference on Artificial Intelligence  
A word sense disambiguation (WSD) system trained on one domain and applied to a different domain will show a decrease in performance. One major reason is the different sense distributions between different domains. This paper presents novel application of two distribution estimation algorithms to provide estimates of the sense distribution of the new domain data set. Even though our training examples are automatically gathered from parallel corpora, the sense distributions estimated are good
more » ... ugh to achieve a relative improvement of 56% when incorporated into our WSD system.
dblp:conf/ijcai/ChanN05 fatcat:d4el2uoxbfegffynf272mslsaq