Researchers have developed and evaluated large language models for assessing suicide risk in Arabic crisis helpline calls, comparing their performance against English translations. The study utilized de-identified transcripts from Lebanon's National Lifeline for Emotional Support and Suicide Prevention, ensuring data privacy by processing audio on-site. Both Arabic and English models were fine-tuned and tested, with the best Arabic model achieving a macro-F1 of 81.19 and ROC-AUC of 90.61 for high-risk calls, while the best English model reached 85.00 and 92.59 respectively. The findings suggest that suicide risk can be effectively classified from de-identified Arabic transcripts, supporting the potential for these models to serve as operator-facing tools. AI
IMPACT Demonstrates LLMs' capability in specialized, privacy-sensitive domains like mental health support, potentially improving crisis intervention efficiency.
RANK_REASON Academic paper detailing novel application of LLMs to a specific domain with performance metrics. [lever_c_demoted from research: ic=1 ai=1.0]
- Arabic
- Columbia Suicide Severity Rating Scale
- English
- large-language models
- Lebanon
- Levantine Arabic
- National Lifeline for Emotional Support and Suicide Prevention
- Transformer encoder
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
