Abstract: With the development of Web, XML has become the de facto standard for data representation and exchange However, a large amount of corporate data resides in RDBMSs Hence, it is inevitable to publish relational data as XML documents and then transfer them to partners At present, widely used publishing is to be done specifically for each customer However, this publishing procedure ignores similarities among publishing requests of the customers, which leads to high publishing cost and long response time Proposed in this paper are algorithms of mining frequent publishing requests and caching intermediate results based on the similarities of publishing requests When a new publishing request comes, the response time can be reduced to a great extent by utilizing the intermediate results Preliminary experiments show that the method is very efficient