Assessing the contribution of shallow and deep knowledge sources for word sense disambiguation

Corpus-based techniques have proved to be very beneficial in the development of efficient and accurate approaches to word sense disambiguation (WSD) despite the fact that they generally represent relatively shallow knowledge. It has always been thought, however, that WSD could also benefit from deeper knowledge sources. We describe a novel approach to WSD using inductive logic programming to learn theories from first-order logic representations that allows corpus-based evidence to be combined with any kind of background knowledge. This approach has been shown to be effective over several disambiguation tasks using a combination of deep and shallow knowledge sources. Is it important to understand the contribution of the various knowledge sources used in such a system. This paper investigates the contribution of nine knowledge sources to the performance of the disambiguation models produced for the SemEval-2007 English lexical sample task. The outcome of this analysis will assist future work on WSD in concentrating on the most useful knowledge sources.