Hindi Word Sense Disambiguation Using Cosine Similarity

Hindi is the regional language of India. Most of the people access, retrieve, and share documents in Hindi language. As all the natural languages possess property of being ambiguous, so does Hindi language, which creates obstacles in usage of information technology properly. In order to remove ambiguity from Hindi language, we need a system called Hindi word sense disambiguation (HWSD). In this paper, we present a supervised method, called HWSD using cosine similarity in which vectors are created for testing query and sense knowledge data for the ambiguous word by considering weights. Experiment is performed on dataset consisting of 90 Hindi ambiguous words and it is found that this method outperforms Lesk's algorithm which is well known algorithm for Word sense disambiguation (WSD). We obtained an overall average precision of 78.99 % and average recall of 72.58 %.