An Improved HowNet-based Algorithm for Semantic Similarity Computation

Abstract

Similarity technology has been widely used in many applications. However, most of the similarity researches focus on English language. In this paper, we develop the improved similarity technology and focus on its application to Chinese language. We use HowNet, which is a Chinese version of WordNet, as the bases. We analyze the structure and data resource of HowNet, and develop an improved information content model of concepts. Meanwhile, we extend the synonyms of concepts, and use them to develop a better similarity computation technique. The experiment results show that our methods achieve the best compared with other existing methods in the degree of similarity accuracy.