A Topical/Local Classifier for Word Sense

TLC is a supervised training (S) system that uses a Bayesian statistical model and features of a word's context to identify word sense. We describe the classifier's operation and how it can be configured to use only topical context cues, only local cues, or a combination of both. Our results on Senseval's final run are presented along with a comparison to the performance of the best S system and the average for S systems. We discuss ways to improve TLC by enriching its feature set and by substituting other decision procedures for the Bayesian model. Future development of supervised training classifiers will depend on the availability of tagged training data. TLC can assist in the hand-tagging effort by helping human taggers locate infrequent senses of polysemous words.