Cohere has released Cohere Transcribe Arabic, an open-source model designed to address the complexities of Arabic speech recognition. The model is built to handle challenges such as dialect diversity, bilingual conversations, and specialized vocabulary. According to Cohere, it is the most accurate open-source Arabic speech-to-text system available, surpassing other models in key benchmarks. Human evaluations of transcripts on a 1-to-5 scale indicate that it outperforms both Whisper Large V3 and the standard Cohere Transcribe model in overall quality, dialect accuracy, and code-switching capabilities.
The 2-billion-parameter ASR model is available under the Apache 2.0 license and can be accessed via Hugging Face or the Cohere API. Additional benchmarks and examples are provided on the Cohere blog. Cohere claims the model outperforms Whisper Large V3 and the standard Cohere Transcribe model in benchmarks. It is specifically tailored for Arabic speech, addressing the language's unique transcription challenges.
Cohere announced the release of the model with a focus on improving the accuracy and reliability of Arabic speech recognition. The model is intended to support a wide range of applications, from real-time transcription to voice-assisted services in Arabic-speaking regions. Source: thedecoder
