unknown

Medibeng Whisper Tiny is a fine-tuned ASR model designed for translating code-switched Bengali-English conversations into English, primarily in clinical settings. It uses the MediBeng dataset and performs well in transcribing healthcare-specific dialogue. The model has been optimized for multilingual environments but may struggle with strong regional accents or highly specialized medical terms. It demonstrates an improved performance through consistent reduction in training loss and Word Error Rate.

pr0mila-gh0sh/MediBeng-Whisper-Tiny · Hugging Face

The official Open Source daily.dev Squad. 

Open Source

We’re excited to announce the release of Medibeng Whisper Tiny, a fine-tuned version of OpenAI’s Whisper model designed for automatic speech recognition (ASR) and translation of code-switched Bengali-English conversations into English. Specifically tailored for clinical and healthcare settings, this model is capable of accurately transcribing and translating speech that includes a natural mix of Bengali and English—common in multilingual medical environments. Medibeng Whisper Tiny is ideal for tasks such as medical transcription, doctor-patient interaction analysis, and clinical documentation in multilingual contexts. The model is freely available for research and non-commercial use.

<p>We’re excited to announce the release of Medibeng Whisper Tiny, a fine-tuned version of OpenAI’s Whisper model designed for automatic speech recognition (ASR) and translation of code-switched Bengali-English conversations into English. Specifically tailored for clinical and healthcare settings, this model is capable of accurately transcribing and translating speech that includes a natural mix of Bengali and English—common in multilingual medical environments. Medibeng Whisper Tiny is ideal for tasks such as medical transcription, doctor-patient interaction analysis, and clinical documentation in multilingual contexts. The model is freely available for research and non-commercial use.</p>