The need for robust and easily extensible systems for word sense disambiguation coupled with successes in training systems for a variety of tasks using large on-line corpora has led to extensive research into corpus-based ...