Detecting and Visualizing the Dispute Structure of the Replying Comments in the Internet Forum Sites

Abstract

Comparing with the existing web pages, one of the popular features of blogs and the web discussion boards is the capability of the interactive communication among users. In online communities such as web logs or Internet discussion boards, users can read articles, as well as write some comments to the articles to express his/her opinion. These kinds of replying comments become an important means of communication between the author who writes the article and the readers of it. Sometimes, we can find new information that does not appear in the contents of the article by reading comments posted to the article. Also, we can figure out various opinions in comments by reading controversial comments. Popular articles, however, frequently get up to thousands of comments, which is too much to be read in a reasonable time. Especially, to find dispute relations in the comments, we have no alternative but to read all the comments. Although there have been several studies to extract an opinion or a controversy from comments or social networks, most of them tend to be dependent on the language used or the typing errors of the contents. In this reason, we propose a method for extracting the dispute relations from comments and visualizing them including the involved commenters. Since comments written by disputing commenters tend to appear in turns, we consider only the order of commenters to detect pairs of commenters in disputing. So, our method is not affected by the language used nor typos in comments. Also, the dispute relations are visualized by an undirected graph, and it is helpful to grasp the degree of controversy intuitively. According to the experimental results, our method is able to detect dispute couples of commenters about 79% on average. Also, we could find unusual commenters such as spammers or bursty commenters as well as a structure of controversy in comments.