Our work in AI for Mental Health (AI-MH) bridges the gap between raw social/conversational data and clinical utility through two major pillars: Foundation Modeling and Resource Building.
Starting from the release of MentalBERT—the first domain-specific pre-trained model for mental healthcare—our mission is to build robust systems and datasets that support early intervention and counseling, providing healthcare professionals with computational tools for more efficient assessment.
Core Research Pillars
Domain-Specific Foundations
Pre-training language models on massive-scale mental health corpora to capture the nuanced, colloquial language of distress that general models often miss.
Emotion-Aware Analysis
Leveraging LLMs with emotion-enhanced prompting strategies to better understand the subjective experiences of individuals in supportive conversations.
Resources and Benchmarks
Constructing datasets and benchmarks for emojis, psychological defenses, emotion support, psychological counseling, and suicide-related social content.
Key Resources
SWMH Dataset
A comprehensive collection of data from Reddit's SuicideWatch and other mental health subreddits for suicide risk assessment.
MentalBERT Series
Continue-pretrained encoder-only models for mental health based on BERT, RoBERTa and clinical variants.
SQPsych
Structured Questionnaire-based Psychotherapy framework that generates synthetic therapist-client conversations to train counseling chatbots.
Publications
You Never Know a Person, You Only Know Their Defenses: Detecting Levels of Psychological Defense Mechanisms in Supportive Conversations
Hongbin Na, Zimu Wang, Zhaoming Chen, Peilin Zhou, Yining Hua, Grace Ziqi Zhou, Haiyang Zhang, Tao Shen, Wei Wang, John Torous, Shaoxiong Ji, and Ling Chen
SuicidEmoji: Derived Emoji Dataset and Tasks for Suicide-Related Social Content
Tianlin Zhang, Kailai Yang, Shaoxiong Ji, Boyang Liu, Qianqian Xie, and Sophia Ananiadou
Towards Explainable Mental Health Analysis with Large Language Models
Kailai Yang, Shaoxiong Ji, Tianlin Zhang, Qianqian Xie, and Sophia Ananiadou
MentalBERT: Publicly Available Pretrained Language Models for Mental Healthcare
Shaoxiong Ji, Tianlin Zhang, Luna Ansari, Jie Fu, Prayag Tiwari, and Erik Cambria
Natural Language Processing Applied to Mental Illness Detection: A Narrative Review
Tianlin Zhang, Annika Schoene, Shaoxiong Ji, and Sophia Ananiadou