Detecting Cyberbullying in Online Communities

Online communities are platforms enabling their users to interact over the web. In particular, they are popular among adolescents as a tool to discuss topics of mutual interest. However, offending communication is a growing issue in these online environments. In its basic form, the process of sending messages over electronic media to cause psychological damage to a victim is called online harassment. In a more severe form, cyberbullying is the process of sending offending messages several times to the same victim by the same offender. In this work, we propose an approach to detect cyberbullies and their victims. Identifying and aiding victims received only brief attention in existing work. We introduce a harassment graph to capture multiple message exchanges comprising cyberbullying cases. We show that our approach is able to precisely detect cyberbullies and their victims. Additionally, we propose metrics to measure the severity of online harassment and cyberbullying cases in terms of quantitative aspects. In particular, the metrics allow to identify victims of severe cyberbullying cases and might be used as an early indicator to provide fast and selective aid by administrators. We further propose use cases for our approach in online communities to tackle the problem of cyberbullying.