Tugstugi Regional ASR — Whisper Medium

This Whisper ASR model was trained on regional Bengali speech data with 10 dialects from the Ben-10 dataset.

Paper: Are ASR foundation models generalized enough to capture features of regional dialects for low-resource languages? (IJCNLP-AACL 2025)

Code: BengaliAI/reg-speech-aacl (training config in finetuning/)

Leaderboard: Ben-10 Regional ASR Leaderboard

Inference Space hosted at https://huggingface.co/spaces/bengaliAI/regional_bengali-asr_tugstugi_whisper-medium

Citation

@inproceedings{dipto-etal-2025-asr,
    title = "Are {ASR} foundation models generalized enough to capture features of regional dialects for low-resource languages?",
    author = "Dipto, Tawsif Tashwar  and
      Hossain, Azmol  and
      Faruque, Rubayet Sabbir  and
      Hassan, Md. Rezuwan  and
      Fatema, Kanij  and
      Shome, Tanmoy  and
      Naswan, Ruwad  and
      Zihad, Md. Foriduzzaman  and
      Anam, Mohaymen Ul  and
      Tasnim, Nazia  and
      Mahmud, Hasan  and
      Hasan, Md Kamrul  and
      Shawon, Md. Mehedi Hasan  and
      Sadeque, Farig  and
      Reasat, Tahsin",
    booktitle = "Proceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics",
    month = dec,
    year = "2025",
    address = "Mumbai, India",
    publisher = "The Asian Federation of Natural Language Processing and The Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.ijcnlp-short.17/",
    doi = "10.18653/v1/2025.ijcnlp-short.17",
    pages = "178--188",
}