Refining the most frequent sense baseline

We refine the most frequent sense baseline for word sense disambiguation using a number of novel word sense disambiguation techniques. Evaluating on the Senseval-3 English all words task, our combined system focuses on improving every stage of word sense disambiguation: starting with the lemmatization and part of speech tags used, through the accuracy of the most frequent sense baseline, to highly targeted individual systems. Our supervised systems include a ranking algorithm and a Wikipedia similarity measure.