Graph-based user classification for informal online political discourse Malouf and Mullen, 2007
R. Malouf and T. Mullen. Graph-based user classification for informal online political discourse. In Proceedings of the 1st Workshop on Information Credibility on the Web, 2007.
Abstract from paper
With the rise of the interactive “Web2.0” and the increasing tendency of online publications to turn to message-board style reader feedback venues, informal political discourse has become an important feature of the intellectual landscape of the Internet. We consider innate political bias or “sentiment” to be of interest for a variety of reasons, including as a factor in determining the reliability of posters in terms of authority and truthfulness. We describe several experiments in identifying the political orientation of posters in an informal environment. Our results indicate that the most promising approach is to augment text classification methods by exploiting information about how posters interact with each other.
This paper experiments with different methods to estimate the political affiliation of a user in the Politics.com dataset. They found that this is a very hard problem and classical machine learning methods such as Naive Bayes did not have a good performance. Using the observation that users tend to quote users from the opposing affiliation, they were able to improve the classification performance.
Exernal link: