Datasets
Standard Dataset
SOS-HL-1K
- Citation Author(s):
- Submitted by:
- Hongzhi Qi
- Last updated:
- Sun, 06/09/2024 - 05:32
- DOI:
- 10.21227/pyzc-8h56
- Data Format:
- License:
- Categories:
- Keywords:
Abstract
We sourced our data by crawling comments from the “Zoufan” blog within the Weibo social platform. Subsequently, a team of qualified psychologists were enlisted to annotate the data. In our study, strict data preprocessing measures were adopted to protect users’ privacy.
SOS-HL-1K (Suicide Risk Classification)
- Categories: High risk, Low risk
- Number of Samples:
- High risk: 601
- Low risk: 648
- Data Split:
- Training set: 999 samples
- Test set: 250 samples
- Average Number of Words per Post: 47.79
- Labels: Each post is labeled with either 'high risk' or 'low risk'.
If you use this dataset in your research, please cite the following paper:
@misc{qi2023evaluating,
title={Evaluating the Efficacy of Supervised Learning vs Large Language Models for Identifying Cognitive Distortions and Suicidal Risks in Chinese Social Media},
author={Hongzhi Qi and Qing Zhao and Changwei Song and Wei Zhai and Dan Luo and Shuo Liu and Yi Jing Yu and Fan Wang and Huijing Zou and Bing Xiang Yang and Jianqiang Li and Guanghui Fu},
year={2023},
eprint={2309.03564},
archivePrefix={arXiv},
primaryClass={cs.CL}
}