ShobdoSetu: A Data-Centric Framework for Bengali Long-Form Speech Recognition and Speaker Diarization
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910060534300672 |
|---|---|
| author | Sakib, Md. Nazmus Tanvir, Shafiul Ahamed, Mesbah Uddin Mukdho, H. M. Aktaruzzaman |
| author_facet | Sakib, Md. Nazmus Tanvir, Shafiul Ahamed, Mesbah Uddin Mukdho, H. M. Aktaruzzaman |
| contents | Bengali is spoken by over 230 million people yet remains severely under-served in automatic speech recognition (ASR) and speaker diarization research. In this paper, we present our system for the DL Sprint 4.0 Bengali Long-Form Speech Recognition (Task~1) and Bengali Speaker Diarization Challenge (Task~2). For Task~1, we propose a data-centric pipeline that constructs a high-quality training corpus from Bengali YouTube audiobooks and dramas \cite{tabib2026bengaliloop}, incorporating LLM-assisted language normalization, fuzzy-matching-based chunk boundary validation, and muffled-zone augmentation. Fine-tuning the \texttt{tugstugi/whisper-medium} model on approximately 21,000 data points with beam size 5, we achieve a Word Error Rate (WER) of 16.751 on the public leaderboard and 15.551 on the private test set. For Task~2, we fine-tune the pyannote.audio community-1 segmentation model with targeted hyperparameter optimization under an extreme low-resource setting (10 training files), achieving a Diarization Error Rate (DER) of 0.19974 on the public leaderboard, and .26723 on the private test set. Our results demonstrate that careful data engineering and domain-adaptive fine-tuning can yield competitive performance for Bengali speech processing even without large annotated corpora. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_19256 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ShobdoSetu: A Data-Centric Framework for Bengali Long-Form Speech Recognition and Speaker Diarization Sakib, Md. Nazmus Tanvir, Shafiul Ahamed, Mesbah Uddin Mukdho, H. M. Aktaruzzaman Computation and Language Bengali is spoken by over 230 million people yet remains severely under-served in automatic speech recognition (ASR) and speaker diarization research. In this paper, we present our system for the DL Sprint 4.0 Bengali Long-Form Speech Recognition (Task~1) and Bengali Speaker Diarization Challenge (Task~2). For Task~1, we propose a data-centric pipeline that constructs a high-quality training corpus from Bengali YouTube audiobooks and dramas \cite{tabib2026bengaliloop}, incorporating LLM-assisted language normalization, fuzzy-matching-based chunk boundary validation, and muffled-zone augmentation. Fine-tuning the \texttt{tugstugi/whisper-medium} model on approximately 21,000 data points with beam size 5, we achieve a Word Error Rate (WER) of 16.751 on the public leaderboard and 15.551 on the private test set. For Task~2, we fine-tune the pyannote.audio community-1 segmentation model with targeted hyperparameter optimization under an extreme low-resource setting (10 training files), achieving a Diarization Error Rate (DER) of 0.19974 on the public leaderboard, and .26723 on the private test set. Our results demonstrate that careful data engineering and domain-adaptive fine-tuning can yield competitive performance for Bengali speech processing even without large annotated corpora. |
| title | ShobdoSetu: A Data-Centric Framework for Bengali Long-Form Speech Recognition and Speaker Diarization |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.19256 |