Speech-to-Speech Translation on CVSS-C (test)
33.8Fr-En ScoreCVSS-C Supervised Baseline
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CVSS-C Supervised BaselineASR Evaluation=open-source2022.10 | 33.8 | 34.6 | 29.4 | 3.1 | 3.2 | 20.8 | |
| Jia et al.ASR Evaluation=proprietary2022.10 | 32.4 | 33.4 | 23.2 | 3.2 | 2.8 | 19 | |
| Spec-T2Paradigm=Supervised learning + pre-training, Architecture=End-to-end, Source languages={Fr, Es, Ru, Et, Lv}, Target language=En2022.10 | 31.8 | 32.3 | 32.9 | 5.2 | 7.5 | 21.9 | |
| Spec-T2Paradigm=Supervised learning, Architecture=End-to-end, Source languages={Fr, Es, Ru, Et, Lv}, Target language=En2022.10 | 27.4 | 27.7 | 25.4 | 4.1 | 2.5 | 17.4 | |
| Cascaded (ASR→TDN→MT→TTS)Paradigm=Unsupervised learning, Architecture=Cascaded, Config=bilingual2022.10 | 21.6 | 21.2 | 25.3 | 7.2 | 7.7 | 16.6 | |
| Spec-T2Paradigm=Unsupervised learning, Architecture=End-to-end, Source languages={Fr, Es, Ru, Et, Lv}, Target language=En2022.10 | 21.2 | 20.1 | 19.9 | 3.2 | 2.8 | 13.4 |