EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hugging Face Blog·5d ago·5 min read
Introducing Falcon ASR
TL;DR
Falcon-ASR is a 1.6 billion parameter speech recognition model for Arabic, particularly the Emirati dialect, achieving a word error rate of 20.92%. It outperforms existing models and supports multiple languages, including English, French, Spanish, and Portuguese.
✦ Why It Matters
Explore Falcon-ASR's capabilities today by testing it with your own audio recordings.
Key Takeaways
- Test Falcon-ASR with your audio for accurate transcriptions.
- Utilize its support for multiple languages without needing a language flag.
- Leverage word-level timestamps for precise audio analysis.
- Consider Falcon-ASR for applications requiring dialectal Arabic recognition.
Full Summary
Falcon-ASR, developed by the Technology Innovation Institute in Abu Dhabi, is a state-of-the-art speech recognition model designed for Arabic, with a focus on the Emirati dialect. It has 1.6 billion parameters and achieved an average word error rate (WER) of 20.92% across six Arabic test sets, outperforming the previous best result of 23.17%.
In internal evaluations specifically for Emirati speech, it recorded a WER of 22.73% and a character error rate (CER) of 10.19%, demonstrating superior transcription accuracy. The model was trained on a diverse dataset that includes various Arabic dialects and English, allowing it to handle everyday speech variations.
Additionally, it provides word-level timestamps for transcriptions, linking each word to its audio position. Falcon-ASR also supports transcription in English, French, Spanish, and Portuguese using the same model weights, making it versatile for multilingual applications.