Mining Patterns of Cyberbullying on Twitter

Cyberbullying refers to the use of text, images, audio and video to harass or harm individuals or groups on a repetitive and non-stop basis in online social networks. The phenomenon has emerged as a serious societal and public health problem that demands accurate methods for the detection of cyberbullying instances to mitigate the consequences. We perform a detailed analysis of a large-scale real-world dataset to identify online social network topology structure features that are the most prominent in enhancing the accuracy of state-of-the-art classification methods for cyberbullying detection. We derive a small subset of features that are fast to compute while differentiating between "normal" users, cyberbullies and victims. Our findings have important implications for the design of future cyberbullying detection schemes.