Word Sense Disambiguation Using Supervised Learning

Abstract

Word sense disambiguation (WSD) holds significant importance within the field of natural language processing which involves the process of determining the accurate interpretation or meaning of a word within a specific context. Supervised learning has been widely used for WSD, where a labeled dataset is used to train a model to predict the sense of an ambiguous word in context. However, the accuracy of the WSD system can be improved by combining multiple approaches. This project proposes a novel approach that combines the mutual information(MI) statistical method and neural network classifier to improve the accuracy of WSD for the English language. Our system leverages the complementary strengths of these two approaches, with mutual information identifying the most informative features for distinguishing between different word senses and neural networks learning complex non-linear relationships between input features and target labels. We evaluate our approach on the Senseva1-2 benchmark corpus and demonstrate that our method outperforms existing state-of-the-art WSD methods. Our approach can potentially improve the accuracy of WSD systems. It can be applied to various NLP tasks where word sense disambiguation is required, such as machine translation, text classification, and information retrieval.