Speech Recognition on MuAViC (test)
197.9Arabic ScoreWhisper large-v2
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Whisper large-v2Pretrain data=680k hrs, 100+ langs, SNR=02025.12 | 197.9 | 244.4 | 113.3 | 116.3 | 172.3 | 172.4 | 223.6 | 126.2 | 170.8 | |
| u-HuBERTPretrain data=1.7k hrs, English, SNR=02025.12 | 102.3 | 73.2 | 69.7 | 43.7 | 43.2 | 48.5 | 47.6 | 67 | 61.9 | |
| XLS-R 300MPretrain data=1.2k hrs, 9 langs, SNR=02025.12 | 97.3 | 69.8 | 74.8 | 47.6 | 37.1 | 47.9 | 54.4 | 59.8 | 61.1 | |
| XLAVS-R 2BPretrain data=1.2k hrs, 9 langs, SNR=02025.12 | 93.5 | 58.5 | 38.6 | 23.9 | 23.5 | 24.6 | 26.1 | 41 | 41.2 | |
| mAV-HuBERT + Hard RoutingPretrain data=7.0k hrs, 100+ langs, SNR=02025.12 | 93.4 | 49.3 | 35.7 | 20.3 | 23.6 | 23.4 | 24.1 | 44.7 | 39.3 | |
| MoHAVE-LargePretrain data=7.0k hrs, 100+ langs, SNR=02025.12 | 92.9 | 47.3 | 35.3 | 18.7 | 21.2 | 21.6 | 21.9 | 40.6 | 37.4 | |
| XLAVS-R 300MPretrain data=8.3k hrs, 100+ langs, SNR=02025.12 | 91.9 | 53.5 | 49.6 | 28.8 | 29.3 | 32.2 | 32.5 | 46.1 | 45.5 | |
| mAV-HuBERT (re-implementation)Pretrain data=7.0k hrs, 100+ langs, SNR=02025.12 | 88.7 | 51.3 | 37.2 | 20.7 | 22.6 | 24.2 | 23.8 | 42.4 | 38.9 | |
| mAV-HuBERTPretrain data=1.7k hrs, English, SNR=02025.12 | 82.2 | 66.9 | 62.2 | 40.7 | 39 | 44.3 | 43.1 | 43.1 | 52.7 |