Skip to content

Dataset · HumynLabs

Noisy medical audio for robustness tests

HumynLabs/medical-opinion-english-audio

Short English audio clips featuring medical complaints with intentionally low quality for robustness testing.

What's inside

An English audio dataset (<1K examples) where participants speak a phrase containing a medical complaint. The data is intentionally collected as 'B-grade': noisy, imperfect, and suboptimal. Designed to evaluate models under degraded input conditions.

What you can do with it

  • Evaluating ASR and audio classifiers on noisy medical recordings
  • Auditing model robustness against degraded audio
  • Testing low-quality audio data filtering
Tasks
Audio classification
Languages
English
Size
up to 1K rows
Downloads this month
272
Likes
543

How to load it

from datasets import load_dataset

ds = load_dataset("HumynLabs/medical-opinion-english-audio")