Rank Computation Methods for Web Documents

We analyze the existing rank computation methods for the Web using a notion of rank flows and focusing on the methods for authority rank computation. We then propose a new model, called the voting model, for computing various ranks of Web pages that can be used by a search engine to better estimate the quality and relevance of the Web pages. The proposed model has a number of advantages over the surfing model used in the Google search engine [4]: it subsumes the surfing model providing natural trivially converging computation of more different types of ranks, as well as easier extensibility, faster personalization, and finer control over its parameters, due to its conceptual simplicity. The voting model also provides a new meaning for the parameters and the results of the Google's surfing model.