Interested in Contributing?

Scored Systems

System Submitter System Notes Constraint Run Notes BLEU BLEU-cased TER BEER 2.0 CharactTER
GTCOM-Primary  (Details) peterzong
GTCOM
transformer, back-translation and clean data by language model and translation models, ensemble decoding. Fine-Tuning with newstest2017 back translation. yes transformer, back-translation and clean data by language model and translation models, ensemble decoding. Fine-Tuning with newstest2017 back translation.

43.8

43.8

failed

0.390

0.522

Alibaba-ensemble-system-with-reranking  (Details) Alibaba_MT
Alibaba-MT
Ensemble of multiple finetuned transformer models + reranking yes

43.5

43.4

failed

0.387

0.545

Alibaba-General-System  (Details) AlibabaMT
Alibaba-MT
transformer, fine-tuning, ensemble, re-ranking no

43.2

43.2

failed

0.386

0.536

NiuTrans  (Details) NiuTrans
Northeastern University
Ensemble of 15 Transformer models with re-ranking yes Ensemble of 15 Transformer models with re-ranking

43.2

43.2

failed

0.389

0.530

Tencent ensemble system  (Details) Mr Translator
Tencent
Rerank 72 ensemble outputs with 48 features (including t2t R2l, t2t L2R, rnn L2R, rnn R2L etc.) Back translation. Joint train with Chinese to English. Fine-tuning with selected data. Ensemble Learning. yes

43.2

43.2

failed

0.388

0.521

Alibaba-General-System  (Details) AlibabaMT
Alibaba-MT
transformer, fine-tuning, ensemble, re-ranking no

43.2

43.1

failed

0.385

0.537

NiuTrans  (Details) NiuTrans
Northeastern University
Ensemble of 15 Transformer models with re-ranking yes Ensemble of 15 Transformer models with re-ranking

43.1

43.1

failed

0.389

0.530

Alibaba-ensemble-model  (Details) Alibaba_MT
Alibaba-MT
Ensemble of multiple finetuned Transformer models yes

43.1

43.1

failed

0.386

0.541

NiuTrans  (Details) NiuTrans
Northeastern University
Ensemble of 15 Transformer models with re-ranking yes Ensemble of 15 Transformer models with re-ranking

42.8

42.8

failed

0.387

0.535

Uni-NMT-Transformer-EnZh  (Details) Unisound
Unisound AI Labs
BackTranslation + Ensemble + Rerank yes BackTranslation + Ensemble + Rerank(EnZh-L2R+EnZh-R2L+ZhEn-L2R+ZhEn-L2R, average rescore weight)

42.7

42.5

failed

0.382

0.541

Tencent single system  (Details) Mr Translator
Tencent
Single LSTM systems with 6 layers encoders and 3 layers decoders. Back translation (ensemble) with parallel target parts and 20 million mono lingual corpus. Trained with R2l regulazition and target to source joint training. Fine-tuning with selected data from CNN and RNN classification model. yes

42.0

42.0

failed

0.381

0.533

lyg5623  (Details) lyg5623
transformer yes youdao

41.2

41.2

failed

0.378

failed

base  (Details) Translator
yes

40.6

40.6

failed

0.372

0.564

Yingjun Jiang  (Details) wangwei
yes

40.6

40.6

failed

0.360

0.561

NICT  (Details) rui.wang
NICT
The same team with benjamin.marie of NICT. yes Transformer, back-translation, ensemble, and reranking

39.7

39.7

failed

0.369

0.574

bit_1  (Details) bit-nmt
first yes

39.5

39.5

failed

0.365

0.581

TencentFmRD-enzh  (Details) Bojie Hu
TencentFmRD
Transformer, ensemble, reranking, finetune, back-translation yes Transformer, ensemble, reranking, finetune, back-translation

39.4

39.4

failed

0.371

0.634

transformer  (Details) weijia
University of Maryland
yes ensemble of 3 transformer models

39.0

39.0

failed

0.340

0.590

Wonder Woman  (Details) fansiawang
Personal
yes beam=12 rerank

38.8

38.8

failed

0.361

0.589

Wonder Woman  (Details) fansiawang
Personal
yes no rerank

38.8

38.8

failed

0.365

0.588

bit_1  (Details) bit-nmt
first yes

38.6

38.6

failed

0.360

0.594

bit_1  (Details) bit-nmt
first yes

38.2

38.2

failed

0.362

0.576

bit_1  (Details) bit-nmt
first yes

37.1

37.1

failed

0.357

0.579

test  (Details) rnn-lstm-gru
no

35.3

35.3

failed

0.342

0.610

test  (Details) rnn-lstm-gru
no

35.3

35.3

failed

0.216

0.731

test  (Details) rnn-lstm-gru
no

35.1

35.1

failed

0.178

0.777

bit_1  (Details) bit-nmt
first yes

34.7

34.7

failed

0.343

0.617

lyg5623  (Details) lyg5623
transformer yes final

33.5

33.5

failed

0.336

0.651

uedin-nmt-2018  (Details) rsennrich
University of Edinburgh
XapaJIaMnu's system with fixed tokenization. yes

33.3

33.3

failed

0.337

0.685

test  (Details) rui.wang
NICT
no

33.0

33.0

failed

0.339

0.614

lyg5623  (Details) lyg5623
transformer yes readstring

32.9

32.9

failed

0.334

0.657

UEDIN  (Details) XapaJIaMnu
UEDIN
yes Best deep with layer normalization and multi-head attention. Small vocabulary (18k) + ensembles trained for about 40 hours

32.1

32.1

failed

0.177

0.807

transfomer  (Details) xxy
basic transformer no

29.8

29.8

failed

0.315

0.744

King_Arthur  (Details) King_Arthur
no the third upload with constraint

19.3

19.3

failed

0.163

failed

King_Arthur  (Details) King_Arthur
no the second submit

18.0

18.0

failed

0.162

failed

King_Arthur  (Details) King_Arthur
no this is the first upload version

17.9

17.9

failed

0.162

1.058

  (Details) neteaseAI
netease
no

16.5

16.5

failed

0.247

1.065

  (Details) neteaseAI
netease
no

16.3

16.3

failed

0.241

0.893

  (Details) neteaseAI
netease
no

16.3

16.3

failed

0.241

0.893

transfomer  (Details) xxy
basic transformer no

failed

failed

failed

0.315

0.744

  (Details) neteaseAI
netease
no

failed

failed

failed

0.157

0.911

test_xzhi  (Details) zhixuhao
Tencent
transformer yes first try

failed

failed

failed

0.344

0.627

  (Details) neteaseAI
netease
no

failed

failed

failed

0.241

0.893