With many security events taking place in recent years, scientists have realized vulnerability management is an important field and it brings critical effects to many information systems and services. One topic in this field is to identify similarity relationship between vulnerabilities. It can help us to alarm potential attacks. In this paper, we propose a text mining approach to compute a similarity score between two vulnerabilities based on their text description. It consists of two steps: preprocessing and similarity score computation. Experimental results based on an annotated vulnerability dataset have proved the effectiveness of our approach.