Merging Word Senses

WordNet, a widely used sense inventory for Word Sense Disambiguation(WSD), is often too fine-grained for many Natural Language applications because of its narrow sense distinctions. We present a semi-supervised approach to learn similarity between WordNet synsets using a graph based recursive similarity definition. We seed our framework with sense similarities of all the word-sense pairs, learnt using supervision on humanlabelled sense clusterings. Finally we discuss our method to derive coarse sense inventories at arbitrary granularities and show that the coarse-grained sense inventory obtained significantly boosts the disambiguation of nouns on standard test sets.