Word similarity computation based on HowNet

Word similarity computation has important applications in many areas, such as natural language processing, intelligence information retrieval, document clustering, automatic answering, word sense disambiguation, machine translation etc. This article makes an intensive study of word similarity computation based on HowNet and the word similarity is computed in three steps: (1) computes sememes similarity, (2) computes concepts similarity using the weight sum method of sememes similarity, (3) takes the maximum similarity of concepts as words similarity. This article mainly introduces numbering the sememes and presents a more accurate method of decomposing the concept description type. The experiment shows that the algorithm of word similarity presented in this article is simple and feasible, and the computed results are relatively consistent with awareness.