⚕️💬 ReDSM5: A Reddit Dataset for DSM-5 Depression Detection

ℹ️ Looking for a quick preview?
A fully paraphrased, anonymized sample with 25 entries is publicly available on the Hugging Face Hub — no user agreement required!

🚦 Access Conditions

This dataset is gated. To obtain access, please complete the access request form at ReDSM5 Agreement Form and submit it via email to [email protected]. Your request will be reviewed and you’ll receive approval or further instructions by email.

📝 Dataset Summary

ReDSM5 is a corpus of 1,484 Reddit posts, each sentence-level annotated for presence or absence of the nine DSM-5 major depressive episode symptoms (or SPECIAL_CASE for expert discrimination cases), along with a clinical rationale from a licensed psychologist.

A public paraphrased sample is available for inspection and prototyping at irlab-udc/redsm5-sample.
This sample includes 25 entries, has been completely paraphrased to protect anonymity, and can be accessed without any agreement.

Unlike other datasets, every annotation includes:

• sentence_text: The relevant sentence from a Reddit post.
• DSM5_symptom: One of:
  • DEPRESSED_MOOD
  • ANHEDONIA
  • APPETITE_CHANGE
  • SLEEP_ISSUES
  • PSYCHOMOTOR
  • FATIGUE
  • WORTHLESSNESS
  • COGNITIVE_ISSUES
  • SUICIDAL_THOUGHTS
  • SPECIAL_CASE (for non-DSM-5 clinical/positive discriminations)
• status:
  • 1 = relevant evidence for the symptom
  • 0 = explicit negative (clinician-annotated absence of symptom)
• explanation: Short clinical rationale.
• post_id, sentence_id: For reference and grouping.

There is also a full post file with the original post text.

📁 Files

• redsm5_posts.csv
  • Columns: post_id, text (cleaned full post)
• redsm5_annotations.csv
  • Columns: post_id, sentence_id, sentence_text, DSM5_symptom, status, explanation

Example annotation row:

post_id	sentence_id	sentence_text	DSM5_symptom	status	explanation
s_3019_239	s_3019_239_1	However my sex drive went crazy...	SPECIAL_CASE	1	The above statement allows us to affirm that the person...
s_427_110	s_427_110_1	I'm feeling pretty proud of myself...	SPECIAL_CASE	1	The person who writes this post is manifesting positive feelings...
s_221_12	s_221_12_5	I have trouble sleeping every night	SLEEP_ISSUES	1	This statement shows persistent sleep issues matching DSM-5 criteria.

📊 Dataset Statistics

• Total posts: 1,484
• Total number of expert explanations: 1,547
• Average explanations per post: 1.04
• Average number of symptoms per post (status=1 only): 1.39
• Number of hard negatives (posts with no symptoms, status=1): 392
• Average post length (in words): 294.7
• Min / Max post length (in words): 2 / 6,990

Symptom Distribution (posts with status=1):

Symptom	Posts tagged
DEPRESSED_MOOD	328
ANHEDONIA	124
APPETITE_CHANGE	44
SLEEP_ISSUES	102
PSYCHOMOTOR	35
FATIGUE	124
WORTHLESSNESS	311
COGNITIVE_ISSUES	59
SUICIDAL_THOUGHTS	165
SPECIAL_CASE	92

📦 File Format

redsm5_full_posts.csv

post_id	text
s_1001_1	I feel tired every day. I can't concentrate and sleep is a mess...

redsm5_annotations_long.csv

post_id	sentence_id	sentence_text	DSM5_symptom	status	explanation
s_1001_1	s_1001_1_1	I feel tired every day.	FATIGUE	1	Indicates chronic fatigue.
s_1001_1	s_1001_1_3	I can't concentrate...	COGNITIVE_ISSUES	1	Persistent cognitive issues.

💡 Notes

• Each row in redsm5_annotations_long.csv is a unique (post, sentence, symptom) expert-annotated evidence.
• For multi-label tasks, group by post_id.
• For post-level text, use redsm5_full_posts.csv.

📝 Citation

This paper has been accepted as a Resource Paper at CIKM 2025. The official conference proceedings will be available soon. In the meantime, you can read the preprint on arXiv:

@misc{bao2025redsm5,
  title        = {ReDSM5: A Reddit Dataset for DSM-5 Depression Detection},
  author       = {Eliseo Bao and Anxo Pérez and Javier Parapar},
  year         = {2025},
  eprint       = {2508.03399},
  archivePrefix= {arXiv},
  primaryClass = {cs.CL},
  url          = {https://arxiv.org/abs/2508.03399},
  note         = {Accepted at CIKM 2025}
}

📬 Contact

For questions, please reach out via email: [email protected]