zenitsu09's picture
Add dataset README with documentation
83a5c26 verified
# Read Speech Test Set
This folder contains the **read speech test set** for the MADASR2.0 ASRU 2025 Challenge.
- Includes ~3 hours of read-speech per language for **8 languages**: `bh`, `bn`, `ch`, `kn`, `mg`, `mr`, `mt`, `te`
- Covers **all 33 dialects** and is **balanced across dialects, domains, speakers, and gender**
## Aggregate Statistics
| Metric | Value |
|--------|-------|
| Total Utterances | 19838 |
| Total Duration (sec) | 108160.46 |
| Average Duration per Utterance (sec) | 5.45 |