Hate Speech Detection

Hate speech detection studies how to identify harmful language and infer who it targets, what it means in context, and how severe it is. Common approaches range from lexicons and supervised classifiers to multilingual transformers and large language models; multimodal systems combine text with images, memes, audio, or video, sometimes using retrieval, reasoning, or explanatory evidence. Evaluation increasingly tests cross-platform and cross-language transfer, implicit hate, code-mixed language, and human disagreement rather than relying only on standard benchmark scores. A central challenge is balancing reliable detection with fairness and low false-positive rates, since sarcasm, reclaimed language, cultural context, and differences in annotator judgment can make harmful intent difficult to label consistently.