What pushes their buttons? Predicting comment polarity from the content of political blog posts

Political blogs as a form of social media allow for an uniquely interactive form of political discourse. This is especially evident in focused blogs with a strong ideological identity. We investigate techniques to identify topics within the context of the community, which when discussed in a blog post evoke a discernible positive or negative collective opinion from readers who respond to posts in comments. This is done by using computational methods to assign sentiment polarity to blog comments and learning community specific models that summarize issues tackled by blogs and predict the polarity based on the topics discussed in a blog pos