Journal Homepage: www.ijcst.org

Abstract — A novel approach to identify the sense of the word appearing in a sentence is proposed in this paper. The basic idea is to facilitate the natural language processing through the broad manifold of unsupervised observations. The target word is tokenized and a relationship with other words is constructed through mapping to constitute the training set using specific domain wordnet. It constitutes the corpus from where similar word senses are arrived at and there from be-hives its role into the question answering system. The scheme is evaluated through precise performance indices for two specific domains to illustrate its applicability in the present day context