The massive heterogeneous Web information resources flood on the Internet, and the Web pages classified have no relation with each other. In this paper, the topic knowledge repository is built to find the semantic relation between Web pages, and the similar relation and the associated relation are defined to describe the semantic relation, which helps to provide knowledge service for user and other services. Web pages are represented by vector space model and classified by the sample Web pages which are chosen previously, through the topic knowledge repository, the semantic relation between Web pages can be found. In the clothing industry, an improved classification algorithm is proved to be much better via the experiment, and a method proposed to find the semantic relation between Web pages is proved feasible.