Tokenizer Transplantation: Mitigating Autoregressive Collapse in Edge-Efficient Bengali ASR

By Sanjid Hasan · Paper · cs.CL

Lightweight speech recognition models are critical for edge deployment, yet highly optimized architectures like Moonshine often fail on morphologically rich, non-Latin languages such as Bengali. This study identifies the root cause of this failure as the model's English-centric b

Cs.cl

View original

HomeResourceLoading…