A Reference Dataset for Training Interventions in Online Hate Speech

Addressing online hate speech is a crucial but complex challenge, one that can be supported through Natural Language Processing (NLP) techniques. While previous research has primarily focused on developing NLP methods to automatically detect hate speech, it has largely overlooked the need for proactive intervention to discourage future use. Additionally, most existing hate speech datasets treat each post as an isolated instance, failing to consider the broader conversational context. In this paper, we introduce a novel task: generative hate speech intervention.