NICT_LOGO.JPG KYOTO-U_LOGO.JPG

WAT

The Workshop on Asian Translation
Evaluation Results

[EVALUATION RESULTS TOP] | [BLEU] | [RIBES] | [AMFM] | [HUMAN (WAT2019)] | [HUMAN (WAT2018)] | [HUMAN (WAT2017)] | [HUMAN (WAT2016)] | [HUMAN (WAT2015)] | [HUMAN (WAT2014)] | [EVALUATION RESULTS USAGE POLICY]

BLEU


# Team Task Date/Time DataID BLEU
Method
Other
Resources
System
Description
juman kytea mecab moses-
tokenizer
stanford-
segmenter-
ctb
stanford-
segmenter-
pku
indic-
tokenizer
unuse myseg kmseg
1ORGANIZERJPCja-en2016/11/16 11:06:501338---49.35--- 0.00 0.00 0.00NMTYesOnline A (2016/11/14)
2JAPIOJPCja-en2017/07/28 22:22:151574---49.00--- 0.00 0.00 0.00NMTYesCombination of 3 NMT systems (OpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor)
3KNU_HyundaiJPCja-en2019/07/27 12:27:323189---48.71------NMTYesTransformer Base (+ASPEC data), relative position, BT, r2l reranking, checkpoint ensemble
4JAPIOJPCja-en2017/07/29 10:49:011578---48.08--- 0.00 0.00 0.00NMTYesOpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor
5sarahJPCja-en2019/07/26 11:18:482969---47.07------NMTNoTransformer, ensemble of 4 models
6ORGANIZERJPCja-en2018/08/15 18:38:511965---44.08---- 0.00 0.00NMTNoNMT with Attention
7JAPIOJPCja-en2017/07/25 18:17:301455---44.07--- 0.00 0.00 0.00NMTNoOpenNMT(dbrnn)
8bjtu_nlpJPCja-en2016/08/16 12:34:361149---41.62--- 0.00 0.00 0.00NMTNoRNN Encoder-Decoder with attention mechanism, single model
9CUNIJPCja-en2017/07/31 22:34:511666---38.29--- 0.00 0.00 0.00SMTNoBahdanau (2014) seq2seq with conditional GRU on byte-pair encoding
10u-tkbJPCja-en2017/07/26 12:53:501472---37.31--- 0.00 0.00 0.00NMTNoNMT with SMT phrase translation (phrase extraction with branching entropy; attention over bidirectional LSTMs; by Harvard NMT)
11NICT-2JPCja-en2016/08/05 17:58:401103---36.06--- 0.00 0.00 0.00SMTYesPhrase-based SMT with Preordering + Domain Adaptation (JPC and ASPEC) + Google 5-gram LM
12ORGANIZERJPCja-en2016/07/26 10:15:251035---35.77--- 0.00 0.00 0.00OtherYesOnline A (2016)
13NICT-2JPCja-en2016/08/04 17:26:271080---35.68--- 0.00 0.00 0.00SMTNoPhrase-based SMT with Preordering + Domain Adaptation
14ORGANIZERJPCja-en2016/07/13 17:12:09980---34.40--- 0.00 0.00 0.00SMTNoString-to-Tree SMT
15Kyoto-UJPCja-en2016/07/27 17:15:101057---33.85--- 0.00 0.00 0.00EBMTNoKyotoEBMT 2016 w/o reranking
16ORGANIZERJPCja-en2016/07/13 17:00:31979---32.23--- 0.00 0.00 0.00SMTNoHierarchical Phrase-based SMT
17ORGANIZERJPCja-en2016/07/13 16:54:09977---30.80--- 0.00 0.00 0.00SMTNoPhrase-based SMT
18ORGANIZERJPCja-en2016/08/05 15:18:461090---21.57--- 0.00 0.00 0.00OtherYesRBMT A (2016)
19ORGANIZERJPCja-en2016/08/05 14:51:271088---21.00--- 0.00 0.00 0.00OtherYesRBMT C (2016)
20ORGANIZERJPCja-en2016/08/05 15:59:141095---18.38--- 0.00 0.00 0.00OtherYesRBMT B (2016)
21ORGANIZERJPCja-en2016/07/26 13:43:211051---16.00--- 0.00 0.00 0.00OtherYesOnline B (2016)

Notice:

Back to top

RIBES


# Team Task Date/Time DataID RIBES
Method
Other
Resources
System
Description
juman kytea mecab moses-
tokenizer
stanford-
segmenter-
ctb
stanford-
segmenter-
pku
indic-
tokenizer
unuse myseg kmseg
1KNU_HyundaiJPCja-en2019/07/27 12:27:323189---0.879063------NMTYesTransformer Base (+ASPEC data), relative position, BT, r2l reranking, checkpoint ensemble
2ORGANIZERJPCja-en2016/11/16 11:06:501338---0.878342---0.0000000.0000000.000000NMTYesOnline A (2016/11/14)
3JAPIOJPCja-en2017/07/28 22:22:151574---0.878298---0.0000000.0000000.000000NMTYesCombination of 3 NMT systems (OpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor)
4JAPIOJPCja-en2017/07/29 10:49:011578---0.873093---0.0000000.0000000.000000NMTYesOpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor
5sarahJPCja-en2019/07/26 11:18:482969---0.870997------NMTNoTransformer, ensemble of 4 models
6JAPIOJPCja-en2017/07/25 18:17:301455---0.863385---0.0000000.0000000.000000NMTNoOpenNMT(dbrnn)
7ORGANIZERJPCja-en2018/08/15 18:38:511965---0.859168----0.0000000.000000NMTNoNMT with Attention
8bjtu_nlpJPCja-en2016/08/16 12:34:361149---0.851975---0.0000000.0000000.000000NMTNoRNN Encoder-Decoder with attention mechanism, single model
9u-tkbJPCja-en2017/07/26 12:53:501472---0.841136---0.0000000.0000000.000000NMTNoNMT with SMT phrase translation (phrase extraction with branching entropy; attention over bidirectional LSTMs; by Harvard NMT)
10CUNIJPCja-en2017/07/31 22:34:511666---0.837425---0.0000000.0000000.000000SMTNoBahdanau (2014) seq2seq with conditional GRU on byte-pair encoding
11NICT-2JPCja-en2016/08/05 17:58:401103---0.825420---0.0000000.0000000.000000SMTYesPhrase-based SMT with Preordering + Domain Adaptation (JPC and ASPEC) + Google 5-gram LM
12NICT-2JPCja-en2016/08/04 17:26:271080---0.824398---0.0000000.0000000.000000SMTNoPhrase-based SMT with Preordering + Domain Adaptation
13ORGANIZERJPCja-en2016/07/26 10:15:251035---0.803661---0.0000000.0000000.000000OtherYesOnline A (2016)
14Kyoto-UJPCja-en2016/07/27 17:15:101057---0.800841---0.0000000.0000000.000000EBMTNoKyotoEBMT 2016 w/o reranking
15ORGANIZERJPCja-en2016/07/13 17:12:09980---0.793483---0.0000000.0000000.000000SMTNoString-to-Tree SMT
16ORGANIZERJPCja-en2016/07/13 17:00:31979---0.763030---0.0000000.0000000.000000SMTNoHierarchical Phrase-based SMT
17ORGANIZERJPCja-en2016/08/05 14:51:271088---0.755017---0.0000000.0000000.000000OtherYesRBMT C (2016)
18ORGANIZERJPCja-en2016/08/05 15:18:461090---0.750381---0.0000000.0000000.000000OtherYesRBMT A (2016)
19ORGANIZERJPCja-en2016/07/13 16:54:09977---0.730056---0.0000000.0000000.000000SMTNoPhrase-based SMT
20ORGANIZERJPCja-en2016/08/05 15:59:141095---0.710992---0.0000000.0000000.000000OtherYesRBMT B (2016)
21ORGANIZERJPCja-en2016/07/26 13:43:211051---0.688004---0.0000000.0000000.000000OtherYesOnline B (2016)

Notice:

Back to top

AMFM


# Team Task Date/Time DataID AMFM
Method
Other
Resources
System
Description
juman kytea mecab moses-
tokenizer
stanford-
segmenter-
ctb
stanford-
segmenter-
pku
indic-
tokenizer
unuse myseg kmseg
1JAPIOJPCja-en2017/07/28 22:22:151574---0.724710---0.0000000.0000000.000000NMTYesCombination of 3 NMT systems (OpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor)
2ORGANIZERJPCja-en2016/11/16 11:06:501338---0.722590---0.0000000.0000000.000000NMTYesOnline A (2016/11/14)
3JAPIOJPCja-en2017/07/29 10:49:011578---0.715560---0.0000000.0000000.000000NMTYesOpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor
4JAPIOJPCja-en2017/07/25 18:17:301455---0.699930---0.0000000.0000000.000000NMTNoOpenNMT(dbrnn)
5ORGANIZERJPCja-en2018/08/15 18:38:511965---0.699460----0.0000000.000000NMTNoNMT with Attention
6u-tkbJPCja-en2017/07/26 12:53:501472---0.697290---0.0000000.0000000.000000NMTNoNMT with SMT phrase translation (phrase extraction with branching entropy; attention over bidirectional LSTMs; by Harvard NMT)
7bjtu_nlpJPCja-en2016/08/16 12:34:361149---0.690750---0.0000000.0000000.000000NMTNoRNN Encoder-Decoder with attention mechanism, single model
8CUNIJPCja-en2017/07/31 22:34:511666---0.681520---0.0000000.0000000.000000SMTNoBahdanau (2014) seq2seq with conditional GRU on byte-pair encoding
9ORGANIZERJPCja-en2016/07/26 10:15:251035---0.673950---0.0000000.0000000.000000OtherYesOnline A (2016)
10NICT-2JPCja-en2016/08/05 17:58:401103---0.672890---0.0000000.0000000.000000SMTYesPhrase-based SMT with Preordering + Domain Adaptation (JPC and ASPEC) + Google 5-gram LM
11ORGANIZERJPCja-en2016/07/13 17:12:09980---0.672760---0.0000000.0000000.000000SMTNoString-to-Tree SMT
12ORGANIZERJPCja-en2016/07/13 17:00:31979---0.672500---0.0000000.0000000.000000SMTNoHierarchical Phrase-based SMT
13Kyoto-UJPCja-en2016/07/27 17:15:101057---0.672040---0.0000000.0000000.000000EBMTNoKyotoEBMT 2016 w/o reranking
14NICT-2JPCja-en2016/08/04 17:26:271080---0.667540---0.0000000.0000000.000000SMTNoPhrase-based SMT with Preordering + Domain Adaptation
15ORGANIZERJPCja-en2016/07/13 16:54:09977---0.664830---0.0000000.0000000.000000SMTNoPhrase-based SMT
16ORGANIZERJPCja-en2016/08/05 15:18:461090---0.521230---0.0000000.0000000.000000OtherYesRBMT A (2016)
17ORGANIZERJPCja-en2016/08/05 14:51:271088---0.519210---0.0000000.0000000.000000OtherYesRBMT C (2016)
18ORGANIZERJPCja-en2016/08/05 15:59:141095---0.518110---0.0000000.0000000.000000OtherYesRBMT B (2016)
19ORGANIZERJPCja-en2016/07/26 13:43:211051---0.486450---0.0000000.0000000.000000OtherYesOnline B (2016)
20sarahJPCja-en2019/07/26 11:18:482969---0.000000------NMTNoTransformer, ensemble of 4 models
21KNU_HyundaiJPCja-en2019/07/27 12:27:323189---0.000000------NMTYesTransformer Base (+ASPEC data), relative position, BT, r2l reranking, checkpoint ensemble

Notice:

Back to top

HUMAN (WAT2019)


# Team Task Date/Time DataID HUMAN
Method
Other
Resources
System
Description
1sarahJPCja-en2019/07/26 11:18:482969UnderwayNMTNoTransformer, ensemble of 4 models
2KNU_HyundaiJPCja-en2019/07/27 12:27:323189UnderwayNMTYesTransformer Base (+ASPEC data), relative position, BT, r2l reranking, checkpoint ensemble

Notice:
Back to top

HUMAN (WAT2018)


# Team Task Date/Time DataID HUMAN
Method
Other
Resources
System
Description

Notice:
Back to top

HUMAN (WAT2017)


# Team Task Date/Time DataID HUMAN
Method
Other
Resources
System
Description
1JAPIOJPCja-en2017/07/28 22:22:15157468.500NMTYesCombination of 3 NMT systems (OpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor)
2JAPIOJPCja-en2017/07/29 10:49:01157867.000NMTYesOpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor
3CUNIJPCja-en2017/07/31 22:34:51166658.000SMTNoBahdanau (2014) seq2seq with conditional GRU on byte-pair encoding
4u-tkbJPCja-en2017/07/26 12:53:50147251.500NMTNoNMT with SMT phrase translation (phrase extraction with branching entropy; attention over bidirectional LSTMs; by Harvard NMT)

Notice:
Back to top

HUMAN (WAT2016)


# Team Task Date/Time DataID HUMAN
Method
Other
Resources
System
Description
1ORGANIZERJPCja-en2016/11/16 11:06:50133871.500NMTYesOnline A (2016/11/14)
2bjtu_nlpJPCja-en2016/08/16 12:34:36114941.500NMTNoRNN Encoder-Decoder with attention mechanism, single model
3ORGANIZERJPCja-en2016/07/26 10:15:25103532.250OtherYesOnline A (2016)
4NICT-2JPCja-en2016/08/04 17:26:27108025.000SMTNoPhrase-based SMT with Preordering + Domain Adaptation
5NICT-2JPCja-en2016/08/05 17:58:40110324.250SMTYesPhrase-based SMT with Preordering + Domain Adaptation (JPC and ASPEC) + Google 5-gram LM
6ORGANIZERJPCja-en2016/08/05 15:18:46109023.750OtherYesRBMT A (2016)
7ORGANIZERJPCja-en2016/07/13 17:12:0998023.000SMTNoString-to-Tree SMT
8ORGANIZERJPCja-en2016/07/13 17:00:319798.750SMTNoHierarchical Phrase-based SMT

Notice:
Back to top

HUMAN (WAT2015)


# Team Task Date/Time DataID HUMAN
Method
Other
Resources
System
Description

Notice:
Back to top

HUMAN (WAT2014)


# Team Task Date/Time DataID HUMAN
Method
Other
Resources
System
Description

Notice:
Back to top

EVALUATION RESULTS USAGE POLICY

When you use the WAT evaluation results for any purpose such as:
- writing technical papers,
- making presentations about your system,
- advertising your MT system to the customers,
you can use the information about translation directions, scores (including both automatic and human evaluations) and ranks of your system among others. You can also use the scores of the other systems, but you MUST anonymize the other system's names. In addition, you can show the links (URLs) to the WAT evaluation result pages.

NICT (National Institute of Information and Communications Technology)
Kyoto University
Last Modified: 2018-08-02