Co-trained Ensemble Models for Weakly Supervised Cyberbullying Detection

Social media has become an inevitable part of individuals’ social and business lives. Its benefits come with various negative consequences. One major concern is the prevalence of detrimental online behavior on social media, such as online harassment and cyberbullying. In this study, we aim to address the computational challenges associated with harassment detection in social media by developing a machine-learning framework with three distinguishing characteristics. (1) It uses minimal supervision in the form of expert-provided key phrases that are indicative of bullying or non-bullying. (2) It detects harassment with an ensemble of two learners that co-train one another; One learner examines the language content in the message, and the other learner considers the social structure. (3) It incorporates distributed word and graph-node representations by training nonlinear deep models. The model is trained by optimizing an objective function that balances a co-training loss with a weak-supervision loss. We evaluate the effectiveness of our approach using post-hoc, crowdsourced annotation of Twitter data, finding that our deep ensembles outperform previous non-deep methods for weakly supervised harassment detection. We also evaluate on a new benchmark to measure the sensitivity of the detectors to language describing particular social groups.