TA-RAG: Tone-Aware Retrieval-Augmented Generation for Peer-Support Health Communication

The paper introduces TA-RAG, a prompt-based tone-aware retrieval-augmented generation framework designed to enhance peer-support health communication, particularly in sensitive contexts like HIV support. TA-RAG incorporates explicit tone control features—stigma-free rewriting, readability adjustment, recipient adaptation, and empathy rephrasing—without requiring model fine-tuning. Evaluation results demonstrate that these components significantly improve communication quality while maintaining content integrity, highlighting the potential for prompt-based tone management in RAG systems for health-related applications.