Machine Translation Human Evaluation on BMELD En-Zh (test)
0.69CoherenceScheduled Multi-task Learning (SML)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Scheduled Multi-task Learning (SML)Model setting=Base, Training framework=Three-stage2022.05 | 0.69 | 0.735 | |
| M-NCTModel setting=Base, Training framework=Three-stage2022.05 | 0.665 | 0.695 | |
| CSA-NCTModel setting=Base, Training framework=Three-stage2022.05 | 0.65 | 0.68 | |
| NCTModel setting=Base, Training framework=Three-stage2022.05 | 0.635 | 0.665 | |
| Trans.Model setting=Base, Training framework=Three-stage2022.05 | 0.62 | 0.655 | |
| Trans. w/o FTModel setting=Base, Training framework=Three-stage2022.05 | 0.585 | 0.63 |