Offline Speech-to-Speech Translation on CVSS-C (test)
84.52Fr-En ASR-BLEUGround Truth
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Ground Truth#Param.=N/A2024.06 | 84.52 | 88.54 | 75.53 | 82.86 | |
| StreamSpeech#Param.=70M, Decoding Strategy=beam102024.06 | 28.45 | 27.25 | 20.93 | 25.54 | |
| UnitY#Param.=67M, Decoding Strategy=beam102024.06 | 27.77 | 24.95 | 18.74 | 23.82 | |
| StreamSpeech#Param.=70M, Decoding Strategy=greedy2024.06 | 27.58 | 26.16 | 19.72 | 24.49 | |
| UnitY#Param.=67M, Decoding Strategy=greedy2024.06 | 26.9 | 23.93 | 18.19 | 23.01 | |
| Translatotron 2#Param.=87M, Decoding Strategy=beam102024.06 | 26.07 | 22.93 | 16.91 | 21.97 | |
| Translatotron 2#Param.=87M, Decoding Strategy=greedy2024.06 | 25.49 | 22.35 | 16.24 | 21.36 | |
| DASpeech#Param.=93M, Decoding Strategy=Viterbi (greedy)2024.06 | 25.03 | 21.37 | 16.14 | 20.85 | |
| S2UT#Param.=73M, Decoding Strategy=beam102024.06 | 22.23 | 18.53 | 2.99 | 14.58 | |
| S2UT#Param.=73M, Decoding Strategy=greedy2024.06 | 20.91 | 16.94 | 2.46 | 13.44 | |
| Translatotron#Param.=79M, Decoding Strategy=greedy2024.06 | 16.96 | 8.72 | 1.97 | 9.22 |