User Classification For Informal Online Politcal Discourse

Google Tech Talk January 16, 2007 ABSTRACT Weblogs, message boards, chat rooms logs, and other informal web texts are an incredibly rich knowledge base. However, the nature of these texts often makes them unsuitable for analysis using many standard natural language processing methods. This talk will describe on-going work in classification of participants in informal on-line political discourse. Lexically-oriented methods, such as sentiment analysis and text classification, have worked well in other domains but perform poorly on this task. We have achieved better results by augmenting these methods with social network analysis, exploiting information about the structure of the on-line community...