HuggingFace, IISc partner to supercharge model building on India's diverse languages

Hugging Face · 1 year ago · 49

The Indian Institute of Science and Hugging Face partnered to make Vaani, an open-source dataset covering India's languages and dialects, more accessible to global AI developers. Vaani targets collection of over 150,000 hours of speech data from 1 million people across all 773 districts, with Phase 1 covering 80 districts already open-sourced and Phase 2 expanding to 100 additional districts. The partnership aims to enable development of AI systems for speech recognition, language identification, and multilingual applications tailored to India's linguistic diversity.