2 papers
cs.CL2026
Tokenizer Transplantation: Mitigating Autoregressive Collapse in Edge-Efficient Bengali ASR
Sanjid Hasan, Md. Abdur Rahman
Lightweight speech recognition models are critical for edge deployment, yet highly optimized architectures like Moonshine often fail on morphologically rich, non-Latin languages su…
cs.SD2026
Make It Hard to Hear, Easy to Learn: Long-Form Bengali ASR and Speaker Diarization via Extreme Augmentation and Perfect Alignment
Sanjid Hasan, Risalat Labib, A H M Fuad +1
Although Automatic Speech Recognition (ASR) in Bengali has seen significant progress, processing long-duration audio and performing robust speaker diarization remain critical resea…