Discovering political tendency in bulletin board discussions by social community analysis

Abstract

Bulletin Board System (BBS) is very popular and provide an asynchronous, text-based environment for users to exchange information and idea. A BBS consists of a number of discussion boards, each of which focuses on a particular subject. A discussion on a topic consists of a seed articles followed by some articles responsive to the seed article or other responsive articles. This paper investigates the social community analysis technique to discover the political tendency of users within the boards from discussions. We first extract the social interactions between users, such as "reply" and "advocate" of posts between users. A social network among users is constructed based on the extracted social interaction. After building the social network, we employ the graph partition, graph coloring, and graph clustering algorithms respectively to discover the social communities. Users of the same community have more potential of political opinion agreement with each other. By using this approach, we are able to partition users into two opposite groups and identify their political tendency effectively without linguistic analysis of discussion content.