A New Approach to Word Sense Disambiguation in MT System

The word sense disambiguation (WSD) is a tough issue in natural language processing. This paper introduces and contrasts the main approaches of WSD prevailing in the world, and analyzes their advantages and disadvantages briefly. Then the author focuses on the vector space model (VSM), and furthermore, puts forward a new method that uses the approach of multi-level sentence similarity (MLSS) computation in the VSM. The new method improves the accuracy of VSM method and overcomes the "bag of words" problem in VSM.