Exploring feature set combinations for WSD

This paper explores the split of features sets in order to obtain better wsd systems through combinations of classifiers learned over each of the split feature sets. Our results show that only k-nn is able to profit from the combination of split features, and that simple voting is not enough for that. Instead we propose combining all k-nn subsystems where each of the k neighbors casts one vote. We have performed a thorough evaluation on two datasets (Senseval-3 Lexical-Sample and All-words), having set the best combination options in a development dataset (Senseval-2 Lexical-Sample). The results for the All-Words task are the best published up to date. The results for the lexical sample are state-of-the-art.