Sarvam AI releases new dataset for testing speech recognition in 22 Indian languages
Sarvam AI has teamed up with AI4Bharat to launch Indic DiarBench. This open benchmark dataset is designed to assess automatic speech recognition (ASR) and speaker diarisation for all 22 scheduled languages in India. The collection includes 108 hours of authentic, multi-speaker dialogue involving 485 different speakers from 189 districts, covering both rural and urban areas. It is now accessible via Hugging Face.
by shortkt.com✓