The existing semantic similarity calculation method of Chinese words, basing on HowNet, cannot be used well for the similarity calculation of fuzzy concept in the specified significance and cannot accurately express the implicit semantic similarity which is vague and in the specified case. For this problem, on the basis of HowNet, combined with label semantic framework, a method based on mass semantic similarity is proposed. This paper presents an algorithm to calculate the semantic distance of Chinese words by using mass function, and proves that this method which is used to calculate the similarity of Chinese words in specified case can effectively improve the accuracy of word similarity calculation, and solve the problem of semantic similarity of Chinese words in specified domain.