SGER: Algorithms and Software Tools for Discovering Coalitions and Identifying Leaders in the Blogosphere

Rensselaer Polytechnic Institute, Troy NY

Investigators

Abstract

Modern means of communication, such as web-logs, chatrooms, and e-mail messaging, provide unprecedented opportunities for a vast number of people to express their opinions and communicate with other people, most of whom are not personal acquaintances. On August 26, 2005, TheWall Street Journal published an article by Daniel Henninger, "A Quite Majority Replaces Vietnam's 'Silent Majority'." He writes: ". . . Richard Nixon . . . gave a famous speech in 1969. . . . The idea, of course, was that the daily attention commanded by the anti-war movement was missing a class of Americans who sat home seething at the behavior of protesters. Today, because of the Internet, no one has to seethe in silence, as wired activists in both parties proved in 2004's high-tech elections, and now. But it may be that the current infatuation with anti-Bush, anti-Iraq sentiment is again missing a political current flowing beneath the surface of the news, just as the media missed . . . the value voters in the 2004 election." The political and social reality of our time includes a huge communication space, "the electronic society," which allows an ever growing number of people to express themselves and actively participate in the social life of the general society. It is expected that drastically increased ways of self-expression and communication should intensify the processes of forming and evolving social trends. The electronic space is so large that it demands sophisticated computer algorithms in order to address important questions and do meaningful analysis. In any social network, links are formed and broken in a never ending dynamics that represents the evolution of the social network. As the network evolves, social groups form and evolve, and occasionally a social group may grow to a size large enough to represent a new movement or coalition. The goal of this research is to track and discover such coalitions as they form and evolve in the Blogosphere. While technically it is easy to track the evolution of one particular actor of an electronic society, it is much more important and much more difficult to discover groups composed of actors that are holding "reasonably" close although not identical opinions, and to differentiate them from those with substantially different opinions. In addition to the immense amount of data to be processed, the researcher must be aware of the volatility of such groups, whose members may agree on some issues, and disagree on others. The straightforward attempt to discover and characterize these groups might involve listing out topics and positions of the actors. This approach is practically impossible not only because of the time necessary to read (by humans) the publications, but also because of the difficulty in characterization: many nuances and subtleties may look contradictory and inconsistent. Using an automated system for reading would not be practical either as it would also require a substantial linguistic analysis of each article.