Introducing Falcon ASR

Summary

The Technology Innovation Institute in Abu Dhabi released Falcon-ASR, a 1.6 billion parameter speech recognition model focused on Arabic, especially the Emirati dialect, while also supporting English, French, Spanish, and Portuguese. On the Open Universal Arabic ASR Leaderboard, Falcon-ASR achieved a 20.92% average word error rate across six test sets, outperforming the best published average of 23.17% in the reference snapshot. The model recorded a 22.73% word error rate and 10.19% character error rate on an internal Emirati evaluation, 4.07% better in word error rate than the next best system compared. Training data included background noise, overlapping speech, music, reverberation, telephony effects, speed, and pitch variations to handle diverse recording conditions. Falcon-ASR builds on the Falcon3-Audio architecture and provides word-level timestamps linking transcribed words to audio positions.

Takeaways

1. 1

   Falcon-ASR achieves a 20.92% average word error rate across six Arabic test sets on the Open Universal Arabic ASR Leaderboard, compared to the best published 23.17%. Its internal Emirati evaluation shows 22.73% WER, 4.07% lower than the next best system tested.

2. 2

   A single set of weights handles Arabic, English, French, Spanish, and Portuguese without a language flag. Training data includes Emirati, Modern Standard Arabic, other Gulf and Arabic dialects, and English.

3. 3

   Training augmentation covered background noise, overlapping speech, music, reverberation, telephony effects, and speed/pitch variation to improve robustness across recording conditions.

FAQ