Abstract: Query rewriting is a basic problem in database research. It is closely related to the query optimizing, datawarehousing, data integrating, semantic caching, etc. At present, the data on the Internet is mostly the semi structured data , and data integration produces a lot of semi structured data views. Thus how to use semi structured views to rewrite the query so as to shorten the respond time receives more attention. And this problem is essentially an NP problem. Here a novel query rewriting approach is put forward. Compared with the traditional approaches, this approach prunes the exponential results of query rewriting plan space by using the knowledge of characteristic of semi structured data and the relation between the subgoals in the query, while the soundness and completeness of algorithm are maintained. Theoretical analysis proves that the approach greatly reduces the cost of algorithm.