Machine Translation on NIST MT 06 2006 (test)
49.98BLEUAdvAug
Evaluation Results
| Method | Links | |
|---|---|---|
| AdvAugLoss Config.=L_aut + L_adv, Extra corpora=true2020.06 | 49.98 | |
| AdvAugLoss Config.=L_aut + L_adv, Extra corpora=false2020.06 | 49.26 | |
| AdvAugLoss Config.=L_clean + L_adv, Extra corpora=false2020.06 | 47.89 | |
| Cheng et al.Loss Config.=not elaborated, Extra corpora=false2020.06 | 46.95 | |
| AdvAugLoss Config.=L_aut, Extra corpora=false2020.06 | 46.73 | |
| Sennrich et al.Loss Config.=not elaborated, Extra corpora=true2020.06 | 46.39 | |
| Edunov et al.Loss Config.=not elaborated, Extra corpora=true2020.06 | 46.2 | |
| Sato et al.Loss Config.=not elaborated, Extra corpora=false2020.06 | 45.75 | |
| Miyato et al.Loss Config.=not elaborated, Extra corpora=false2020.06 | 45.28 | |
| AdvAugLoss Config.=L_mixup, Extra corpora=false2020.06 | 45.12 | |
| Vaswani et al.Loss Config.=L_clean, Extra corpora=false2020.06 | 44.57 | |
| RNNSEARCH + D_RNN# Para.=+0.17M, Speed=1.2K2018.11 | 37.54 | |
| RNNSEARCH-COVERAGE + D_RNN + O_CDR# Para.=+1.20M, Speed=0.7K2018.11 | 37.42 | |
| RNNSEARCH + D_CNN# Para.=+0.23M, Speed=1.0K2018.11 | 37.36 | |
| RNNSEARCH + O_CDR# Para.=+0M, Speed=1.0K2018.11 | 37.34 | |
| RNNSEARCH + MRT_BLEU# Para.=+0M, Speed=0.3K2018.11 | 37.22 | |
| RNNSEARCH + D_RNN + O_CDR# Para.=+0.17M, Speed=0.8K2018.11 | 36.87 | |
| RNNSEARCH-COVERAGE# Para.=+1.03M, Speed=1.2K2018.11 | 36.52 | |
| RNNSEARCH + O_BLEU# Para.=+0M, Speed=1.1K2018.11 | 36.21 | |
| MemDecType=Attention-based Seq2Seq with external memory2018.08 | 35.98 | |
| RNNSEARCH# Para.=86.7M, Speed=1.4K2018.11 | 35.98 | |
| Seq2Seq + SACTMechanism=Self-Adaptive Control of Temperature2018.08 | 35.95 | |
| RNNSEARCH + O_CHRF3# Para.=+0M, Speed=1.1K2018.11 | 35.95 | |
| CoverageType=Attention-based Seq2Seq with coverage model2018.08 | 34.25 | |
| Seq2Seq2018.08 | 33.54 | |
| RNNSearchType=Attention-based Seq2Seq2018.08 | 31.61 | |
| MosesType=Phrase-based translation, Language Model=4-gram2018.08 | 30.81 |