A New Approach to Word Sense Disambiguation Based on Context Similarity

This paper presents a WSD approach that is based on inner product of vectors algorithm, a supervised approach in which sense-tagged data is used to train the classifier and the results are promising compared to the methods existing in the literature.

Abstract

One of the major issues in the process of machine translation is word sense disambiguation (WSD), which is defined as choosing the correct meaning of a multi-meaning word. Supervised learning methods are usually used to solve this problem. The disambiguation task is carried out using the statistics of the translated documents (as training data) or dual corpora of source and target languages. In this paper we present a supervised learning method for WSD, which is based on Cosine Similarity. As the first step, we extract two sets of features; the set of words that have occurred frequently in the text and the set of words surrounding the ambiguous word. We will present the results of evaluating the proposed schemes and illustrate the effect of weighting strategies proposed. The results are promising compared to the methods existing in the literature. In corpora-based Translation methods translations are generated on the basis of statistical or probabilistic models whose parameters are extracted from the analysis of a bilingual corpus. Statistical translation is based on the study of frequencies of various linguistic units, including words, lexemes, morphemes, letters, etc., in a sample corpus in order to calculate a set of probabilities, so that various linguistic problems such as ambiguity can be solved. In This paper, we present a WSD approach that is based on inner product of vectors algorithm. The proposed scheme is a supervised approach in which sense-tagged data is used to train the classifier.