Dataset Viewer
The dataset viewer is not available for this dataset.
Unexpected token '<', "<html> <h"... is not valid JSON

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Reddit Mental Health Risk Classification Dataset

Dataset Summary

This dataset is a curated mental health text classification dataset designed for training Natural Language Processing (NLP) models to identify different levels of mental health risk from social media posts.

The dataset was created by combining and preprocessing two publicly available Kaggle datasets:

  1. Sentiment Analysis for Mental Health
  2. Suicidal Tweet Detection Dataset

The original datasets were cleaned, filtered, relabeled, and merged to create a unified three-class mental health risk classification dataset.


Dataset Sources

1. Sentiment Analysis for Mental Health

Source: https://www.kaggle.com/datasets/suchintikasarkar/sentiment-analysis-for-mental-health

This dataset contains mental health-related statements labeled with multiple mental health categories including:

  • Depression
  • Anxiety
  • Stress
  • Bipolar Disorder
  • Personality Disorder
  • Suicidal
  • Normal

It aggregates and cleans data from several publicly available mental health datasets, making it suitable for research and machine learning applications.


2. Suicidal Tweet Detection Dataset

Source: https://www.kaggle.com/datasets/aunanya875/suicidal-tweet-detection-dataset

This dataset consists of annotated tweets for suicidal ideation detection and binary classification of suicidal versus non-suicidal content.


Dataset Construction

The final dataset was created through the following pipeline:

  1. Combined both source datasets.
  2. Removed duplicate samples.
  3. Removed missing and invalid records.
  4. Standardized text formatting.
  5. Unified label schema.
  6. Mapped the original labels into three risk categories.

Label Mapping

The final dataset contains three classes:

Label Risk Level Description
0 High Risk Explicit suicidal ideation, suicide planning, self-harm intent, or immediate crisis.
1 Low Risk General discussions, emotionally neutral posts, or non-crisis mental health conversations.
2 Moderate Risk Depression, anxiety, emotional distress, hopelessness, stress, or substance abuse without explicit suicidal intent.

Dataset Statistics

  • Total Samples: 47,894
  • Number of Classes: 3
Class Samples
High Risk 12,064
Moderate Risk 19,487
Low Risk 16,343

Dataset Format

The dataset consists of two columns:

Column Description
statement Text of the social media post
status Integer class label

Example:

statement status
I don't think I can keep fighting anymore. 0
I've been feeling anxious for months. 2
I had a wonderful day with my friends. 1

Intended Uses

This dataset is intended for:

  • Mental health risk classification
  • NLP research
  • Text classification
  • Transformer model fine-tuning
  • Mental health AI research
  • Educational and academic purposes

Limitations

  • The dataset is derived from publicly available social media text and may not represent the broader population.
  • Labels are intended for research purposes and should not be interpreted as clinical diagnoses.
  • Models trained on this dataset should not be used as the sole basis for medical or emergency decisions.

Ethical Considerations

This dataset contains discussions related to depression, anxiety, emotional distress, and suicidal ideation. It is intended exclusively for research, educational, and machine learning purposes.

Users should ensure that any models trained using this dataset are deployed responsibly and include appropriate human oversight for real-world applications.


Citation

If you use this dataset in your research, please cite both the original dataset creators and this processed version.

Original Datasets


License

Please refer to the licenses of the original Kaggle datasets. This repository contains a processed and relabeled version of publicly available data intended for research and educational use.


Acknowledgements

This dataset was prepared as part of the Google Summer of Code (GSoC) 2026 project under HumanAI. We gratefully acknowledge the creators of the original Kaggle datasets for making their work publicly available to support mental health research and NLP development.

Downloads last month
54