Datasets:
Great ASR resource hoping for Bangladeshi Bangla TTS datasets next!
#2
by Sanzid - opened
Great resource for building robust ASR models in Bangladeshi Bangla! A major challenge in Bengali NLP/Speech is that most available open-source datasets heavily represent Indian/Kolkata Bangla dialects and accents. While SUBAK.KO helps bridge that gap for speech-to-text, the community still faces a severe shortage of dedicated, studio-quality Bangladeshi TTS datasets. Hope to see researchers prioritize TTS datasets next to complete the ecosystem.