SAHSD: Enhancing Hate Speech Detection in LLM-Powered Web Applications via Sentiment Analysis and Few-Shot Learning (ACM Digital Library)

Due to skewed training data and a lack of annotated samples, hate speech moderation has become a major concern as large language models (LLMs) increasingly power online apps. Sentiment-Aided Hate Speech Detection (SAHSD), which is presented in this paper, uses sentiment analysis to increase the accuracy of pre-trained models. By combining sentiment and hate speech cues, SAHSD improves detection through few-shot learning and trains a sentiment model using public datasets. SAHSD’s promise for safer, more responsible online moderation is demonstrated by experiments on SBIC and HateXplain, which demonstrate that it performs better than sophisticated techniques and surpasses GPT-4 in generalization.

https://dl.acm.org/doi/10.1145/3696410.3714644