Inherit/Feedback: A New Web Topic-Specific Mining Method

Classical hyperlink analysis algorithms （such as PageRank, HITS） focus on the authority of Web page rather than its topic Thus the crawler based on these algorithms would rapidly drift away in the course of crawling In this paper a new hyperlink analysis method called Inherit/Feedback is presented The key idea is that a page inherits the topic specific correlation from its ancestors and gets the feedback from its descendants There are various applications that can be enhanced by the Inherit/Feedback method, such as pages ranking and topic specific crawling A new topic specific crawling algorithm based on Inherit/Feedback （IFC） is also proposed The experiments show that IFC performs quite well while guiding the topic specific crawling agent and it can be applied to the further discovery and mining from topic specific website