Using Random walks for mining web document associations

Abstract

World Wide Web has emerged as a primetry means for storing and structuring information. In this paper, we present a framework for mining implicit associations among Web documents. We focus on the following problem: “For a given set of seed URLs, find a list of Web pages which reflect the association among these seeds.” In the proposed framework, associations of two documents are induced by the connectivity and linking path length. Based on this framework, we have developed a random walk-hased Web mining technique and validated it by experiments on real Web data. In this paper, we also discuss the extension of the algorithm for considering document contents.