Classification of rumors and counter-rumors

Rumors are posing a serious threat in this digital era. They spread fast and wide on social media to become viral. While some users are deceived by rumors, others explicitly seek to debunk them. In consequence, two opposite forces of claims, namely, rumors and counter-rumors emerge. Given that rumors and counter-rumors are antithetical and could exhibit contrary patterns, the objective of this paper is to classify the two. For this purpose, messages on a rumoring phenomenon were collected from Twitter, and a set of content- based as well as user-based features were identified to distinguish between rumors and counter-rumors. To ensure methodological rigor, a set of classification algorithms including voting was employed using five-fold cross validation. Furthermore, feature selection techniques were used to identify the top features in classifying rumors and counter-rumors. The implications of the results are discussed.