Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

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

Source

arXiv

arxiv.org/abs/2608.11200v1

Read the source

Source

  • Source name: arXiv
  • Source kind: paper
  • Authors / organization: Chen Lyu, Xingwei Tan, Simon Cullen, Shelley Wilson, Lois Arthurs, Arshad Jhumka, Gabriele Pergola
  • Venue: arXiv: cs.CL
  • Published: 2026-08-11
  • Collected: 2026-08-12
  • URL: http://arxiv.org/abs/2608.11200v1
  • PDF: https://arxiv.org/pdf/2608.11200v1
  • DOI: UNKNOWN
  • Registry group: academic_apis_and_preprints
  • Evidence role: canonical_academic_index
  • Language: en
  • Country / region: UNKNOWN

Plain-Language Brief

The study presents ConVAWG, a framework that generates synthetic dialogues about Violence Against Women and Girls (VAWG) using a retrieval-based approach. The framework constructs scenarios based on persona seeds, demographic patterns, crime definitions, and real cases of Domestic Homicide Review. It then converts these scenarios into multi-scene role-play dialogues with targeted toxicity control. The research includes over 6,000 dialogue events across 200 scenarios with detailed metadata. The results show strong dialogue quality and domain fidelity.

What the Document Says

  • The study introduces ConVAWG, a framework for generating synthetic dialogues about Violence Against Women and Girls (VAWG).
  • ConVAWG constructs scenarios from persona seeds, demographic patterns, crime definitions, and Domestic Homicide Review cases.
  • The framework converts these scenarios into multi-scene role-play dialogues.
  • It applies targeted activation-steered toxicity control to appropriate utterances.
  • The research includes over 6,000 dialogue events across 200 scenarios.

Key Concepts Mentioned

  • Violence Against Women and Girls (VAWG)
  • ConVAWG
  • synthetic dialogue generation
  • retrieval-grounded framework
  • persona seeds
  • demographic patterns
  • official crime definitions
  • Domestic Homicide Review cases
  • multi-turn dialogues
  • temporally unfolding phenomenon
  • toxicity control

Claims Made by the Document

  • The framework generates high-quality and domain-fidelity synthetic dialogues about VAWG.

Evidence Used

  • Extensive human evaluation, LLM-as-Judge assessment, ablations, and downstream tasks show strong dialogue quality and domain fidelity.

Methods / Approach

  • Retrieval-grounded approach
  • multi-turn dialogues
  • targeted activation-steered toxicity control

Limitations Stated by the Authors

  • Privacy and legal constraints make it difficult to release large-scale real conversation datasets.
  • Existing work has mostly focused on sentence-level toxicity of online abuses.

Limitations Noted by the Agent

Agent interpretation — not from the source.

It is unclear whether the generated dialogues fully capture the complexity and diversity of real-world VAWG scenarios due to the limited amount of data used in the study.

Open Questions

  • How well do the generated dialogues reflect real-world VAWG scenarios beyond the provided dataset?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The study focuses on modelling VAWG as multi-turn dialogues, but it does not specify how it handles non-verbal communications or offline abuses that may not be conversationally disclosed.

Raw Abstract / Excerpt

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.