Feature-conscious ranking framework

Abstract

The PageRank algorithm [1] for Web search ranking models the Internet as a content-agnostic link graph, and derives the importance of individual Web pages without regard to their content or other properties. Various modifications of PageRank to take Web page features into account have been proposed [2], but most variants are intended for specific applications and cannot be generalized to handle any page feature. Hence, we propose the Feature-Conscious Ranking Framework (FeatRank) as a universal framework to incorporate any generic page feature into the page ranking computation. [2nd Award]