Authors: Kailai Yang, Tianlin Zhang, Ziyan Kuang, Qianqian Xie, Jimin Huang, Sophia Ananiadou

Published on: September 24, 2023

Impact Score: 8.45

Arxiv code: Arxiv:2309.13567

Summary

• What is new: Combines LLaMA2 foundation models with a new multi-task and multi-source dataset (IMHI) for interpretable mental health analysis on social media.
• Why this is important: Low interpretability in traditional methods and unsatisfactory classification performance of LLMs in zero/few-shot mental health analysis.
• What the research proposes: Building MentalLLaMA, the first open-source LLM series specifically for interpretable mental health analysis, using a newly created IMHI dataset.
• Results: MentalLLaMA approaches the correctness of state-of-the-art methods and generates high-quality explanations.

Technical Details

Technological frameworks used: LLaMA2 foundation models

Models used: ChatGPT, MentalLLaMA

Data used: 105K data samples from 10 sources covering 8 mental health tasks

Potential Impact

Mental health tech startups, social media platforms, healthcare analytics firms

Want to implement this idea in a business?

We have generated a startup concept here: MindfulAI Insights.