An Investigation Into Various Approaches For Bengali Long-Form Speech Transcription and Bengali Speaker Diarization

Researchers developed a multistage AI approach for Bengali speech transcription and speaker diarization, achieving significant improvements in processing long-form audio recordings. The system used fine-tuned Whisper models and custom segmentation techniques to address the low-resource nature of Bengali in speech technology applications.

• →Fine-tuned Whisper Medium model achieved 0.38 WER for Bengali speech transcription on private leaderboard.
• →Custom speaker diarization system achieved 0.27 DER using pyannote integration with specialized segmentation models.
• →Two-pass methodology with hyperparameter tuning proved effective for handling noisy acoustic environments.
• →Research demonstrates potential for improving AI inclusivity for South Asian low-resource languages.
• →Complete implementation code made publicly available for community development and research.