Document-Level Machine Translation on TED English-German (test)
0.2806s-BLEUG-Transformer+mBART
Evaluation results
| Method | Links | ||
|---|---|---|---|
| G-Transformer+mBARTPre-training Setting=mBART2023.05 | 0.2806 | 30.03 | |
| G-Transformer+BERTPre-training Setting=BERT2023.05 | 0.2681 | — | |
| Flat-Transformer+BERTPre-training Setting=BERT2023.05 | 0.2661 | — | |
| G-Transformer (fnt.) + Target-side augmentationInitialization=fine-tuned (fnt.), Context Level=document-level, Data Augmentation=Target-side augmentation2023.05 | 0.2659 | 29.2 | |
| G-Transformer + Back-translation (doc) + Target-side augmentationContext Level=document-level, Data Augmentation=Back-translation + Target-side augmentation2023.05 | 0.2621 | 28.58 | |
| Transformer (sent baseline) + Target-side data augmentationContext Level=sentence-level, Data Augmentation=Target-side augmentation2023.05 | 0.2614 | — | |
| Transformer + Back-translation (sent) + Target-side augmentationContext Level=sentence-level, Data Augmentation=Back-translation + Target-side augmentation2023.05 | 0.2613 | — | |
| RecurrentMem2023.05 | 0.2562 | 29.47 | |
| G-Transformer + Back-translation (doc)Context Level=document-level, Data Augmentation=Back-translation2023.05 | 0.2545 | 28.06 | |
| MultiResolution2023.05 | 0.2524 | 29.27 | |
| G-Transformer (fnt.) (doc baseline)Initialization=fine-tuned (fnt.), Context Level=document-level2023.05 | 0.252 | 27.94 | |
| G-TransformerInitialization=fine-tuned (fnt.)2023.05 | 0.2512 | 27.17 | |
| SMDT2023.05 | 0.2512 | — | |
| Hybrid Context2023.05 | 0.251 | — | |
| Transformer + Back-translation (sent)Context Level=sentence-level, Data Augmentation=Back-translation2023.05 | 0.2503 | — | |
| Transformer (sent baseline)Context Level=sentence-level2023.05 | 0.2491 | — | |
| Flat-Transformer2023.05 | 0.2487 | — | |
| HAN2023.05 | 0.2458 | — | |
| SAN2023.05 | 0.2442 | — | |
| G-TransformerInitialization=random (rnd.)2023.05 | 0.2353 | 25.84 |