Towards generalisable hate speech detection: a review on obstacles and solutions (PeerJ Computer Science)

One form of damaging internet content is hate speech, which targets or incites hatred toward a particular group or its members based on their real or imagined identity, including their sexual orientation, religion, or ethnicity. One area of growing interest in natural language processing is automatic hate speech identification, given the surge in hate speech on the internet. The fact that current models don’t perform well when applied to new data, however, has only lately come to light. In an effort to summarize the generalizability of current hate speech detection models and the reasons behind their limited capacity to do so, this survey study also reviews previous attempts to tackle the primary roadblocks and suggests future research avenues aimed at enhancing generalization in hate speech detection.

https://www.researchgate.net/publication/352490211_Towards_generalisable_hate_speech_detection_a_review_on_obstacles_and_solutions