Speech Translation on MuST-C EN-ES (tst-COMMON)
31BLEUJT Proposed
Evaluation results
| Method | Links | |
|---|---|---|
| JT Proposed#pars(m)=76, Cross-Attentive Regularization=true, Online Knowledge Distillation=true2021.07 | 31 | |
| JT-S-MT + CARCross-Attentive Regularization=true2021.07 | 30.4 | |
| JT-S-MT2021.07 | 29.7 | |
| JT-S-ASR2021.07 | 29.4 | |
| JT#pars(m)=762021.07 | 29 | |
| SpeechformerInference Time=1.3x2021.09 | 28.5 | |
| ST#pars(m)=762021.07 | 28.1 | |
| Inaguma et al.2021.07 | 28 | |
| Inaguma et al.2021.09 | 28 | |
| Our baselineInference Time=1.0x2021.09 | 27.9 | |
| + compressionInference Time=0.9x2021.09 | 27.9 | |
| Plain ConvAttentionInference Time=1.8x2021.09 | 27.7 | |
| Wang et al.2021.09 | 27.2 | |
| Gangi et al.#pars(m)=302021.07 | 20.9 |