The Generalized Web Surfer

Different models have been proposed for improving the results of Web search by taking into account the link structure of the Web. The PageRank algorithm models the behavior of a random surfer alternating between random jumps to new pages and following out links with equal probability. We propose to improve on PageRank by using an intelligent surfer that combines link structure and content to decide on its next transition. The intelligent surfer is guided by the textual authority of the web page. The textual authority gives a non-topical estimate of the intrinsic quality of a web page, and when combined with the link-based social authority gives a more complete and robust estimate of the document authoritativeness. Experiments on a number of queries indicate that our algorithm significantly outperforms PageRank algorithms in the human-rated quality of the pages returned while retaining the efficiency and topic-independence characteristics of PageRank.