ABSTRACT

WSD is an essential process in linguistics because of its requirement for the applications based on languages. The first attempt for automated word sense disambiguation was modelled in the context of Machine translation. Many works have been performed on Word sense disambiguation for English, IIT Bombay performed WSD for Hindi language using WordNet. Most of the Natural languages have multiple word ambiguity. Unlike previous approaches which assume single word ambiguity, we used a graph based Unsupervised WSD approach which disambiguates words (with multiple ambiguities) in the sentence. In the above mentioned method, initially, a semantic graph is built for each and every interpretation which is possible. To construct a graph, we will use WordNet for the Hindi Language developed at IIT Bombay. Then we will try to obtain the spanning tree with the minimum cost. The tree must correspond to the interpretation graph and the graph with the minimum interpretation is identified. If the value of the cost found is below the threshold provided, then the corresponding interpretation would be the resulting interpretation of the sentence. This process is continued until accurate results are achieved. This approach considers some open class words. We are extending it to include pronoun, conjunctions, prepositions etc.