Abstract: The research results of James Martin et al.show that most of software faults come from the requirements phase.To improve the quality of software products,that the software requirements specification (SRS) must maintain consistency,correctness and unambiguous is very important.The functional requirement statements in SRS are made word division and speech tagging by natural language processing technology.Each functional requirement statement is converted to a weight vector based on the key words and the functional requirement statements with similarity semantics can be clustered by grey clustering algorithm.The results of clustering can not only facilitate the requirements analysis to review SRS,but also facilitate the developer and software maintenance to carry out activities.

Key words: Requires,Software requires specification (SRS),Software fault,Clustering