ConVAWG: A Retrieval-Grounded Framework for Generating Synthetic VAWG Dialogues

Researchers have introduced ConVAWG, a retrieval-grounded framework designed to generate controlled synthetic dialogues in the domain of Violence Against Women and Girls (VAWG). The framework addresses the challenge of studying conversational dynamics in sensitive areas where real data are scarce due to privacy and legal constraints. Unlike previous work that focused on sentence-level toxicity of online abuse, ConVAWG models abuse as a relational and temporally unfolding phenomenon through multi-turn dialogues. The system builds scenarios from persona seeds and demographic information, generating CPS-aligned synthetic chat dialogues. This approach enables the study of both online and offline abuse, including threats, coercion, surveillance, isolation, stalking, and physical violence, which may be planned or disclosed in conversations. The framework is detailed in a paper announced on arXiv with identifier 2608.11200, categorized as a cross-announcement. The work aims to fill a gap in modelling abuse in a more comprehensive and realistic manner, providing a valuable resource for researchers in fields such as computational social science, NLP, and gender-based violence studies.