Partitioning and Social Scaling of Political Debates using Signed Bipartite Graphs

Abstract

Blogosphere plays an increasingly important roleas a forum for public debate. In this paper, given amixed set of blogs debating a set of political issuesfrom opposing camps, we use signed bipartite graphsfor modeling debates, and we propose an algorithmfor partitioning both the blogs, and the issues (i.e.topics, leaders, etc.) comprising the debate into binaryopposing camps. Simultaneously, our algorithmscales both the blogs and the underlying issues ona univariate scale. Using this scale, a researchercan identify moderate and extreme blogs within eachcamp, and polarizing vs. unifying issues. Throughperformance evaluations we show that our proposedalgorithm provides an eective solution to the problem,and performs much better than existing baselinealgorithms adapted to solve this new problem. In ourexperiments, we used both real data from politicalblogosphere and US Congress records, as well assynthetic data which were obtained by varying polarizationand degree distribution of the vertices ofthe graph to show the robustness of our algorithm.