ABSTRACT
Word Sense Disambiguation (WSD) is an important problem of Natural Language Processing (NLP) with a motive to find the correct sense of a word in the text. We have proposed a novel approach to use epidemic spreading based model for WSD. Prior work used various graph-based centralities and quoted best results using degree centrality measure. Since, a degree does not provide us with any evident information for the conclusion of a sense. It is not a very good measure to be used. However, Expected Force computes the influence of each node on all other nodes of the network and thus has proved to be a better centrality measure. Experimental results back our claims. Expected Force gave better results than other centrality measures and thus added value to our approach.
