Abstract: Many Web information retrieval methods are related to special Web sites, for example, the method based on extracting rules and the one based on training page samples. These methods can do well in a Web site but fail in the others without adding new rules

Key words: Cluster, Information retrieval, Machine learning, Instance of node