Abstract: With the development of Web 2.0, many websites allow users to create and manage their social tags. A lot of searches show that social annotations can be used to improve search quality, but the real tagging system is often sparse, uncategorized, lack of structure and of low quality, therefore traditional SimRank algorithm is so difficult to work. Introducing Jaccard index to SimRank algorithm, we put forward the improvement of social tagging Jaccard SimRank (JSR) similarity calculation method which automatically analyzes the similarity of user-input social annotations and expands them to increase the density. JSR algorithm can make full use of the information of social tagging to achieve effective retrieval and to describe similarity between any two tags intuitively. The experimental datasets come from bibsonomy website, and we have applied Jaccard index, SimRank and JSR algorithms against the test datasets. Experimental results show that the JSR algorithm is more effective in improving search quality than the traditional algorithms.