ConVAWG: A Retrieval-Grounded Framework for Controlled Synthetic Dialogue Generation in Violence Against Women and Girls

2026-08-11

AI Summary

This study addresses the critical gap in existing research on violence against women and girls (VAWG), which has largely relied on single-sentence toxicity detection and lacks multi-turn dialogue data capable of capturing relational dynamics and temporal evolution. To bridge this gap, the work proposes a retrieval-based, controllable synthesis framework that models VAWG as a multi-turn conversational process. By integrating real-world case reports, character profiles, and legal definitions of offenses, the framework leverages retrieval-augmented generation, role-playing modeling, and a hierarchical event-temporal structure to construct CPS-compliant dialogues across diverse scenarios. It further introduces an activation-guided toxicity control mechanism to ensure contextual appropriateness. The project releases over 6,000 high-quality multi-turn dialogues spanning 200 scenarios, annotated with three-tier metadata, and validated through both human evaluation and LLM-as-Judge assessments for fidelity and domain relevance.

Abstract

Synthetic dialogue generation offers a way to study conversational dynamics in sensitive domains where real data are difficult to access, release, or annotate. The underlying abuse may occur online or offline: threats and coercion can appear directly in messages, while behaviours such as surveillance, isolation, stalking, and physical violence may be planned, disclosed, or referred to conversationally. Privacy and legal constraints make it difficult the release of large-scale real conversation datasets; existing work has mostly focused on sentence-level toxicity of online abuses, leaving a gap in modelling abuse as a relational and temporally unfolding phenomenon. In this work, we focus on modelling Violence Against Women and Girls (VAWG) scenarios as multi-turn dialogues. We introduce ConVAWG, a retrieval-grounded framework for generating CPS-aligned synthetic VAWG chat dialogues. ConVAWG builds scenarios from persona seeds, demographic patterns reported by the UK Office for National Statistics, official crime definitions, and retrieved Domestic Homicide Review cases; converts them into hierarchical event timelines; generates multi-scene role-play dialogues; and applies targeted activation-steered toxicity control to appropriate utterances. We release over 6,000 multi-turn dialogue events across 200 scenarios with rich scenario-, event-, and turn-level metadata. Extensive human evaluation, LLM-as-Judge assessment, ablations, and downstream tasks show strong dialogue quality and domain fidelity.

Problem

• Violence Against Women and Girls
• synthetic dialogue generation
• retrieval-grounded framework
• multi-turn dialogue
• relational abuse modeling

Innovation

• retrieval-grounded generation
• controlled synthetic dialogue
• multi-turn VAWG modeling
• activation-steered toxicity control
• hierarchical event timelines