Detecting Opinionated Claims in Online Discussions

This paper explores the automatic detection of sentences that are opinionated claims, in which the author expresses a belief. We use a machine learning based approach, investigating the impact of features such as sentiment and the output of a system that determines committed belief. We train and test our approach on social media, where people often try to convince others of the validity of their opinions. We experiment with two different types of data, drawn from Live Journal web logs and Wikipedia discussion forums. Our experiments show that sentiment analysis is more important in Live Journal, while committed belief is more helpful for Wikipedia. In both corpora, n-grams and part-of-speech features also account for significantly better accuracy. We discuss the ramifications behind these differences.