A common method of word similarity computing is based on a semantic dictionary. This paper introduces the “HowNet” in the basic concepts and hierarchical architecture, drawing lessons from Liu Qun, Li Sujian basic theory in word similarity. It makes use of “HowNet” sememe hierarchical architecture to calculate sememe similarity, then calculates the similarity of the concept, and finally gets the similarity of the words. This paper also makes appropriate adjustment to the computing method, so that the calculated results can be more in line with the actual situation.