Datasets:
license: cc-by-nc-sa-4.0
task_categories:
- text-classification
language:
- zh
tags:
- mental-health
- social-media
size_categories:
- 1M<n<10M
MATCHA: A Multi-Disease Chinese Mental Health Dataset with Cross-Lingual Analyses
1. Overview / 项目简介
MATCHA (Multi-Disorder mental health daTa in CHinA) is a large-scale, community-driven Chinese Weibo dataset designed for multi-disorder mental health research. It is the first Chinese multi-disease dataset that links community posts from Weibo Super-Topics (超话) with user-level historical timelines.
MATCHA（Multi-Disorder mental health daTa in CHinA）是一个大规模、社区驱动的中文微博数据集，专为多病种精神健康研究设计。该数据集首次将微博超话社区帖子与用户层面的历史时间线相结合，构建了完整的用户数字足迹。
2. Key Features / 核心亮点
- Large Scale / 规模宏大: Covers 7 DSM-5 disorders and a control group, totaling 1.8 million posts from 47k users.
- LLM-Assisted Stratification / 大模型辅助标注: Utilizes Qwen-3 to identify "Identified Positive" (IP) users who have explicitly self-reported their clinical status with >80% accuracy.
- Cross-Lingual Alignment / 跨语言对齐: Fully aligned with the English Twitter-STMHD dataset to facilitate cross-lingual psychiatric analysis.
- Dual-Context / 双重语境: Links disorder-specific community posts (STP) with general homepage posts (HP) to characterize behavioral patterns.
3. Data Statistics / 数据统计
| Disorder / 疾病类型 | User Count / 用户数 | Total Posts / 总帖子数 |
|---|---|---|
| ADHD (多动症) | 197 | 10,084 |
| ASD (孤独症) | 753 | 33,622 |
| Bipolar Disorder (双相) | 612 | 125,653 |
| Depression (抑郁症) | 25,202 | 1,232,940 |
| General Anxiety (焦虑症) | 10,004 | 170,874 |
| Obsessive-Compulsive (强迫症) | 442 | 82,006 |
| Schizophrenia (精分) | 584 | 25,816 |
| Control Group (对照组) | 4,125 | 155,136 |
4. Ethics & Data Access / 伦理与数据获取
- De-identification: All personal identifiers have been removed or hashed to protect user privacy.
- Access Control: The dataset is distributed through a controlled-access system for academic, non-commercial use only.
- Ethics Statement: All data were collected from publicly accessible posts without intruding on restricted content.
5. Access Instructions / 申请须知
To access the MATCHA dataset, you must complete the following Dual-Verification process. Requests without a corresponding signed DUA will be declined.
如需访问 MATCHA 数据集，您必须完成以下双重验证流程。未提交签署协议的申请将被拒绝。
Download & Sign: Visit our GitHub Repository to download and sign the
DUA.mdfile. 下载并签署：前往 GitHub 仓库 下载并签署DUA数据使用协议。Email the DUA: Send the signed document to baihan.li612@gmail.com. 发送邮件：将签署后的扫描件发送至 baihan.li612@gmail.com。
- Subject:
[MATCHA Access Request] Name - Institution
- Subject:
Match Email Address: Ensure your Hugging Face account email is the SAME as the one in your signed DUA for verification. 邮箱一致性：请确保您的 Hugging Face 账号邮箱 与协议中填写的邮箱完全一致，以便核对。
We will review your request once the signed DUA is received via email. 我们在收到邮件及签署协议后将尽快予以通过。