Semantic similarity is widely used in many fields such as information retrieval, information extraction, natural language classification, semantic understanding, and machine translation based on word meaning. Especially in recent years, with the rapid development of computer technology, the means of knowledge structure processing is more complicated and diversified, which promotes the rapid development of semantic similarity calculation methods. Aiming at the advantages and disadvantages of these three traditional methods of semantic similarity calculation, an improved method in the paper is presented by using the distance as the basis of the calculation model and combining the two key decision factors of the content and attributes of the information.