Speech Recognition on EasyCom

Metric: WER (%) (lower is better)

LeaderboardDataset
Loading chart...
#ModelWER (%)Extra DataPaperDateCode
1ReVISE (bf)52.1NoReVISE: Self-Supervised Speech Resynthesis with ...2022-12-21-
2ReVISE (ch2)55NoReVISE: Self-Supervised Speech Resynthesis with ...2022-12-21-
3DAJA (MVDR,HMA,1000) (Overlapped Speech)62.36NoDirection-Aware Joint Adaptation of Neural Speec...2022-07-15-
4Demucs (bf)69.8NoReVISE: Self-Supervised Speech Resynthesis with ...2022-12-21-
5Demucs (ch2)86.8NoReVISE: Self-Supervised Speech Resynthesis with ...2022-12-21-