Machine Translation on NIST Chinese-English MT03-MT06 (test)
51.11Average ScoreSB-NMT
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SB-NMTStrategy=Synchronous bidirectional decoding2019.05 | 51.11 | 51.87 | 51.5 | 51.23 | 49.83 | 3.92 | |
| ABD-NMTBase model=Transformer, Strategy=Asynchronous bidirectional decoding2019.05 | 48.19 | 49.47 | 48.01 | 48.19 | 47.09 | 1 | |
| Rerank-NMTBase model=Transformer, Strategy=k-best reranking2019.05 | 48.1 | 48.23 | 48.91 | 48.73 | 46.51 | 0.91 | |
| TransformerDirection=Left-to-right2019.05 | 47.19 | 47.63 | 48.32 | 47.51 | 45.31 | — | |
| Transformer (R2L)Direction=Right-to-left2019.05 | 46.11 | 46.79 | 47.01 | 46.5 | 44.13 | -1.08 | |
| RNMTType=RNN-based NMT2019.05 | 42.47 | 42.43 | 44.56 | 41.94 | 40.95 | -4.72 | |
| SAWRBackbone=Transformer, Heads=8, Layers=6, Hidden dimension=5122019.05 | 41.78 | — | — | — | — | 1.04 | |
| Tree-RNNBackbone=Transformer, Heads=8, Layers=6, Hidden dimension=5122019.05 | 41.42 | — | — | — | — | 0.68 | |
| Tree-LinearizationBackbone=Transformer, Heads=8, Layers=6, Hidden dimension=5122019.05 | 41.26 | — | — | — | — | 0.52 | |
| TransformerBackbone=Transformer, Heads=8, Layers=6, Hidden dimension=5122019.05 | 40.74 | — | — | — | — | — | |
| SAWRsyntax-integrated=true2019.05 | 38.83 | — | — | — | — | 1.74 | |
| Tree-RNNsyntax-integrated=true2019.05 | 38.41 | — | — | — | — | 1.32 | |
| Tree-Linearizationsyntax-integrated=true2019.05 | 38.32 | — | — | — | — | 1.23 | |
| MosesType=phrase-based SMT2019.05 | 37.78 | 37.47 | 41.2 | 36.41 | 36.03 | -9.41 | |
| Baseline2019.05 | 37.09 | — | — | — | — | — | |
| Li et al. (2017)2019.05 | 36.15 | — | — | — | — | 1.45 | |
| Chen et al. (2017b)2019.05 | 35.65 | — | — | — | — | 1.52 | |
| Chen et al. (2017a)2019.05 | 34.3 | — | — | — | — | 2.59 |