A Partitioning Algorithm for Bipartite Graphs to Extract Latent Weblog Communities

Abstract

I propose the concept of a latent weblog community (LBC), as a means to promote the autonomous organization of knowledge on the Internet. Such communities can be illustrated in terms of bipartite graphs based on weblog update information, and they can effectively function to create meeting spaces for bloggers who write about similar or closely related topics but do not know each other. To extract these communities from blogspace, I developed a partitioning algorithm known as the Weakest Pair (WP) algorithm, which separates the weakest pairs of bloggers and webpages, respectively, using co-citation information. As a result of numerical evaluation, the WP algorithm is more effective than the Shortest Path Betweenness (SPB) algorithm in terms of information loss and completeness of bipartite graphs. I will provide three examples of LBC extracted using the WP algorithm and report its secondary effects, i.e. personae detection, the detection of a set of weblogs owned by a single blogger.