CRUSH: Contextually Regularized and User anchored Self-supervised Hate speech Detection
This addresses the problem of cyber-bullying and hate speech proliferation on social platforms for users and moderators, representing an incremental advance in detection methods.
The paper tackles hate speech detection on social media by introducing CRUSH, a framework using user-anchored self-supervision and contextual regularization, achieving ~1-12% improvement in test metrics over previous best approaches on multiple English social media datasets.
The last decade has witnessed a surge in the interaction of people through social networking platforms. While there are several positive aspects of these social platforms, the proliferation has led them to become the breeding ground for cyber-bullying and hate speech. Recent advances in NLP have often been used to mitigate the spread of such hateful content. Since the task of hate speech detection is usually applicable in the context of social networks, we introduce CRUSH, a framework for hate speech detection using user-anchored self-supervision and contextual regularization. Our proposed approach secures ~ 1-12% improvement in test set metrics over best performing previous approaches on two types of tasks and multiple popular english social media datasets.