README.md

metadata

license: cc-by-sa-4.0

Context

A stressor is a specific cause or trigger of an individual’s mental health issue, such as stress, anxiety, or suicidal thoughts. Identifying stressors is a key step toward understanding and improving mental well-being.

Content

This dataset contains 4,001 manually annotated online posts collected from June 2018 to June 2022. Each sample includes the following fields:

• text: The content of the individual's post.
• labels: The identified stressor(s) within the post, including their start and end positions (word-level).
• time_interval: A numeric value (1-8) representing the half-year period in which the post was published.
  • For example, 1 indicates the post was published between 2018-07-01 and 2018-12-31,
    while 8 corresponds to 2022-01-01 to 2022-06-30.

Data Partitioning

By default, we divide the data based on time to simulate a real-world scenario: 1. Samples from time_interval = 1 to 7 serve as training data (historical period). 2. Samples from time_interval = 8 serve as test data (most recent half-year). This setup allows researchers to train models on past data and evaluate performance on future data.

However, you are free to re-partition the dataset based on your specific task or experimental design.

Citation

Wang, Xin, et al. "MISE: Meta-knowledge Inheritance for Social Media-Based Stressor Estimation." Proceedings of the ACM on Web Conference 2025.