Skip to main content
Peer-reviewed Authoritative

Classifying the Information Needs of Survivors of Domestic Violence in Online Health Communities Using Large Language Models: Prediction Model Development and Evaluation Study

Collects 294 Reddit posts from women self-identifying as experiencing intimate partner violence, defines eight information-need classes (shelters, legal, police, safety planning, etc.), augments to 2,216 samples with GPT-3.5, and fine-tunes GPT-3.5 for multiclass classification with a per-class training strategy. Reports an F1 of 70.5% on real posts.

Publisher

Journal of Medical Internet Research (JMIR)

Published

12 May 2025

Added

today

Key Findings

  • Fine-tuned GPT-3.5 classified eight domestic-violence information-need categories from forum posts at F1 70.5% (95% CI 60.6-80.4)
  • The fine-tuned model outperformed base GPT-3.5/GPT-4 and a fine-tuned Llama 2-7B
  • Heavy reliance on synthetic augmentation (294 real → 2,216 samples) is a stated limitation

Methodology Notes

Peer-reviewed, Journal of Medical Internet Research 2025;27:e65397 (12 May 2025), DOI 10.2196/65397. Small real-data base augmented with GPT-3.5-generated samples. jmir.org is JS-rendered to fetchers; verified via Crossref and Europe PMC.

Sources

JMIR article (primary)

Europe PMC record

Archived snapshot (Wayback Machine) — preserved against link rot

Authors

Shaowei Guan, Vivian Hui, Gregor Stiglic, Rose Eva Constantino, Young Ji Lee, Arkers Kwan Ching Wong

Tags

jmirdomestic-violenceipvllm-classificationsignpost

Cite This

APA

Shaowei Guan et al. (2025). Classifying the Information Needs of Survivors of Domestic Violence in Online Health Communities Using Large Language Models: Prediction Model Development and Evaluation Study. Journal of Medical Internet Research (JMIR). https://www.jmir.org/2025/1/e65397