Word Sense Disambiguation Using Neural Networks with Concept Co-occurrence Information.

Most previous word sense disambiguation approaches based on neural networks were impractical due to their huge feature set size. We propose a method for resolving word sense ambiguity using neural networks with refined concept co-occurrence information (CCI) as features. Using CCI refinement processing, we reduce the number of features of the network to a practical size. We also show that word sense disambiguation can be improved by combining several clues rather than using them independently. Our method is fully automated and does not require any hand coding of large-scale resources.