WAT

The Workshop on Asian Translation

Evaluation Results

BLEU

#	Team	Task	Date/Time	DataID	BLEU										Method	Other Resources	System Description
#	Team	Task	Date/Time	DataID	juman	kytea	mecab	moses- tokenizer	stanford- segmenter- ctb	stanford- segmenter- pku	indic- tokenizer	unuse	myseg	kmseg	Method	Other Resources	System Description
1	ORGANIZER	JPCja-en	2016/11/16 11:06:50	1338	-	-	-	49.35	-	-	-	0.00	0.00	0.00	NMT	Yes	Online A (2016/11/14)
2	sakura	JPCja-en	2024/08/08 19:55:46	7281	-	-	-	49.30	-	-	-	-	-	-	NMT	No	LLM: Rakuten/RakutenAI-7B-chat Fine-Tuned with JPC Corpus in six direction (En-Ja, Ja-En, Ko-Ja, Ja-Ko, Zh-Ja, Ja-Zh) - Best
3	sakura	JPCja-en	2024/08/09 00:16:39	7293	-	-	-	49.30	-	-	-	-	-	-	NMT	No	LLM: Rakuten/RakutenAI-7B-chat Fine-Tuned with JPC Corpus in six direction (En-Ja, Ja-En, Ko-Ja, Ja-Ko, Zh-Ja, Ja-Zh)
4	JAPIO	JPCja-en	2017/07/28 22:22:15	1574	-	-	-	49.00	-	-	-	0.00	0.00	0.00	NMT	Yes	Combination of 3 NMT systems (OpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor)
5	KNU_Hyundai	JPCja-en	2019/07/27 12:27:32	3189	-	-	-	48.71	-	-	-	-	-	-	NMT	Yes	Transformer Base (+ASPEC data), relative position, BT, r2l reranking, checkpoint ensemble
6	Bering Lab	JPCja-en	2021/04/23 13:05:24	5419	-	-	-	48.44	-	-	-	-	-	-	NMT	Yes	Transformer Ensemble with additional crawled parallel corpus
7	JAPIO	JPCja-en	2017/07/29 10:49:01	1578	-	-	-	48.08	-	-	-	0.00	0.00	0.00	NMT	Yes	OpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor
8	goku20	JPCja-en	2020/09/22 00:08:48	4108	-	-	-	47.34	-	-	-	-	-	-	NMT	No	mBART pre-training transformer, ensemble of 3 models
9	sarah	JPCja-en	2019/07/26 11:18:48	2969	-	-	-	47.07	-	-	-	-	-	-	NMT	No	Transformer, ensemble of 4 models
10	goku20	JPCja-en	2020/09/21 12:15:47	4094	-	-	-	46.71	-	-	-	-	-	-	NMT	No	mBART pre-training transformer, single model
11	tpt_wat	JPCja-en	2021/04/27 02:32:27	5710	-	-	-	45.86	-	-	-	-	-	-	NMT	No	Base Transformer model with separate vocab, size 8k
12	ORGANIZER	JPCja-en	2018/08/15 18:38:51	1965	-	-	-	44.08	-	-	-	-	0.00	0.00	NMT	No	NMT with Attention
13	JAPIO	JPCja-en	2017/07/25 18:17:30	1455	-	-	-	44.07	-	-	-	0.00	0.00	0.00	NMT	No	OpenNMT(dbrnn)
14	bjtu_nlp	JPCja-en	2016/08/16 12:34:36	1149	-	-	-	41.62	-	-	-	0.00	0.00	0.00	NMT	No	RNN Encoder-Decoder with attention mechanism, single model
15	CUNI	JPCja-en	2017/07/31 22:34:51	1666	-	-	-	38.29	-	-	-	0.00	0.00	0.00	SMT	No	Bahdanau (2014) seq2seq with conditional GRU on byte-pair encoding
16	u-tkb	JPCja-en	2017/07/26 12:53:50	1472	-	-	-	37.31	-	-	-	0.00	0.00	0.00	NMT	No	NMT with SMT phrase translation (phrase extraction with branching entropy; attention over bidirectional LSTMs; by Harvard NMT)
17	NICT-2	JPCja-en	2016/08/05 17:58:40	1103	-	-	-	36.06	-	-	-	0.00	0.00	0.00	SMT	Yes	Phrase-based SMT with Preordering + Domain Adaptation (JPC and ASPEC) + Google 5-gram LM
18	ORGANIZER	JPCja-en	2016/07/26 10:15:25	1035	-	-	-	35.77	-	-	-	0.00	0.00	0.00	Other	Yes	Online A (2016)
19	NICT-2	JPCja-en	2016/08/04 17:26:27	1080	-	-	-	35.68	-	-	-	0.00	0.00	0.00	SMT	No	Phrase-based SMT with Preordering + Domain Adaptation
20	ORGANIZER	JPCja-en	2016/07/13 17:12:09	980	-	-	-	34.40	-	-	-	0.00	0.00	0.00	SMT	No	String-to-Tree SMT
21	Kyoto-U	JPCja-en	2016/07/27 17:15:10	1057	-	-	-	33.85	-	-	-	0.00	0.00	0.00	EBMT	No	KyotoEBMT 2016 w/o reranking
22	ORGANIZER	JPCja-en	2016/07/13 17:00:31	979	-	-	-	32.23	-	-	-	0.00	0.00	0.00	SMT	No	Hierarchical Phrase-based SMT
23	ORGANIZER	JPCja-en	2016/07/13 16:54:09	977	-	-	-	30.80	-	-	-	0.00	0.00	0.00	SMT	No	Phrase-based SMT
24	ORGANIZER	JPCja-en	2016/08/05 15:18:46	1090	-	-	-	21.57	-	-	-	0.00	0.00	0.00	Other	Yes	RBMT A (2016)
25	ORGANIZER	JPCja-en	2016/08/05 14:51:27	1088	-	-	-	21.00	-	-	-	0.00	0.00	0.00	Other	Yes	RBMT C (2016)
26	ORGANIZER	JPCja-en	2016/08/05 15:59:14	1095	-	-	-	18.38	-	-	-	0.00	0.00	0.00	Other	Yes	RBMT B (2016)
27	ORGANIZER	JPCja-en	2016/07/26 13:43:21	1051	-	-	-	16.00	-	-	-	0.00	0.00	0.00	Other	Yes	Online B (2016)

Notice:

This table is sorted by the leftmost segmenters. You can change the segmenter used to sort by clicking each segmenter link.

RIBES

#	Team	Task	Date/Time	DataID	RIBES										Method	Other Resources	System Description
#	Team	Task	Date/Time	DataID	juman	kytea	mecab	moses- tokenizer	stanford- segmenter- ctb	stanford- segmenter- pku	indic- tokenizer	unuse	myseg	kmseg	Method	Other Resources	System Description
1	Bering Lab	JPCja-en	2021/04/23 13:05:24	5419	-	-	-	0.879638	-	-	-	-	-	-	NMT	Yes	Transformer Ensemble with additional crawled parallel corpus
2	KNU_Hyundai	JPCja-en	2019/07/27 12:27:32	3189	-	-	-	0.879063	-	-	-	-	-	-	NMT	Yes	Transformer Base (+ASPEC data), relative position, BT, r2l reranking, checkpoint ensemble
3	ORGANIZER	JPCja-en	2016/11/16 11:06:50	1338	-	-	-	0.878342	-	-	-	0.000000	0.000000	0.000000	NMT	Yes	Online A (2016/11/14)
4	JAPIO	JPCja-en	2017/07/28 22:22:15	1574	-	-	-	0.878298	-	-	-	0.000000	0.000000	0.000000	NMT	Yes	Combination of 3 NMT systems (OpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor)
5	sakura	JPCja-en	2024/08/08 19:55:46	7281	-	-	-	0.877741	-	-	-	-	-	-	NMT	No	LLM: Rakuten/RakutenAI-7B-chat Fine-Tuned with JPC Corpus in six direction (En-Ja, Ja-En, Ko-Ja, Ja-Ko, Zh-Ja, Ja-Zh) - Best
6	sakura	JPCja-en	2024/08/09 00:16:39	7293	-	-	-	0.877343	-	-	-	-	-	-	NMT	No	LLM: Rakuten/RakutenAI-7B-chat Fine-Tuned with JPC Corpus in six direction (En-Ja, Ja-En, Ko-Ja, Ja-Ko, Zh-Ja, Ja-Zh)
7	JAPIO	JPCja-en	2017/07/29 10:49:01	1578	-	-	-	0.873093	-	-	-	0.000000	0.000000	0.000000	NMT	Yes	OpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor
8	goku20	JPCja-en	2020/09/22 00:08:48	4108	-	-	-	0.872474	-	-	-	-	-	-	NMT	No	mBART pre-training transformer, ensemble of 3 models
9	sarah	JPCja-en	2019/07/26 11:18:48	2969	-	-	-	0.870997	-	-	-	-	-	-	NMT	No	Transformer, ensemble of 4 models
10	goku20	JPCja-en	2020/09/21 12:15:47	4094	-	-	-	0.869896	-	-	-	-	-	-	NMT	No	mBART pre-training transformer, single model
11	tpt_wat	JPCja-en	2021/04/27 02:32:27	5710	-	-	-	0.866360	-	-	-	-	-	-	NMT	No	Base Transformer model with separate vocab, size 8k
12	JAPIO	JPCja-en	2017/07/25 18:17:30	1455	-	-	-	0.863385	-	-	-	0.000000	0.000000	0.000000	NMT	No	OpenNMT(dbrnn)
13	ORGANIZER	JPCja-en	2018/08/15 18:38:51	1965	-	-	-	0.859168	-	-	-	-	0.000000	0.000000	NMT	No	NMT with Attention
14	bjtu_nlp	JPCja-en	2016/08/16 12:34:36	1149	-	-	-	0.851975	-	-	-	0.000000	0.000000	0.000000	NMT	No	RNN Encoder-Decoder with attention mechanism, single model
15	u-tkb	JPCja-en	2017/07/26 12:53:50	1472	-	-	-	0.841136	-	-	-	0.000000	0.000000	0.000000	NMT	No	NMT with SMT phrase translation (phrase extraction with branching entropy; attention over bidirectional LSTMs; by Harvard NMT)
16	CUNI	JPCja-en	2017/07/31 22:34:51	1666	-	-	-	0.837425	-	-	-	0.000000	0.000000	0.000000	SMT	No	Bahdanau (2014) seq2seq with conditional GRU on byte-pair encoding
17	NICT-2	JPCja-en	2016/08/05 17:58:40	1103	-	-	-	0.825420	-	-	-	0.000000	0.000000	0.000000	SMT	Yes	Phrase-based SMT with Preordering + Domain Adaptation (JPC and ASPEC) + Google 5-gram LM
18	NICT-2	JPCja-en	2016/08/04 17:26:27	1080	-	-	-	0.824398	-	-	-	0.000000	0.000000	0.000000	SMT	No	Phrase-based SMT with Preordering + Domain Adaptation
19	ORGANIZER	JPCja-en	2016/07/26 10:15:25	1035	-	-	-	0.803661	-	-	-	0.000000	0.000000	0.000000	Other	Yes	Online A (2016)
20	Kyoto-U	JPCja-en	2016/07/27 17:15:10	1057	-	-	-	0.800841	-	-	-	0.000000	0.000000	0.000000	EBMT	No	KyotoEBMT 2016 w/o reranking
21	ORGANIZER	JPCja-en	2016/07/13 17:12:09	980	-	-	-	0.793483	-	-	-	0.000000	0.000000	0.000000	SMT	No	String-to-Tree SMT
22	ORGANIZER	JPCja-en	2016/07/13 17:00:31	979	-	-	-	0.763030	-	-	-	0.000000	0.000000	0.000000	SMT	No	Hierarchical Phrase-based SMT
23	ORGANIZER	JPCja-en	2016/08/05 14:51:27	1088	-	-	-	0.755017	-	-	-	0.000000	0.000000	0.000000	Other	Yes	RBMT C (2016)
24	ORGANIZER	JPCja-en	2016/08/05 15:18:46	1090	-	-	-	0.750381	-	-	-	0.000000	0.000000	0.000000	Other	Yes	RBMT A (2016)
25	ORGANIZER	JPCja-en	2016/07/13 16:54:09	977	-	-	-	0.730056	-	-	-	0.000000	0.000000	0.000000	SMT	No	Phrase-based SMT
26	ORGANIZER	JPCja-en	2016/08/05 15:59:14	1095	-	-	-	0.710992	-	-	-	0.000000	0.000000	0.000000	Other	Yes	RBMT B (2016)
27	ORGANIZER	JPCja-en	2016/07/26 13:43:21	1051	-	-	-	0.688004	-	-	-	0.000000	0.000000	0.000000	Other	Yes	Online B (2016)

Notice:

This table is sorted by the leftmost segmenters. You can change the segmenter used to sort by clicking each segmenter link.

AMFM

#	Team	Task	Date/Time	DataID	AMFM										Method	Other Resources	System Description
#	Team	Task	Date/Time	DataID	unuse	unuse	unuse	unuse	unuse	unuse	unuse	unuse	unuse	unuse	Method	Other Resources	System Description
1	JAPIO	JPCja-en	2017/07/28 22:22:15	1574	-	-	-	0.724710	-	-	-	0.000000	0.000000	0.000000	NMT	Yes	Combination of 3 NMT systems (OpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor)
2	ORGANIZER	JPCja-en	2016/11/16 11:06:50	1338	-	-	-	0.722590	-	-	-	0.000000	0.000000	0.000000	NMT	Yes	Online A (2016/11/14)
3	JAPIO	JPCja-en	2017/07/29 10:49:01	1578	-	-	-	0.715560	-	-	-	0.000000	0.000000	0.000000	NMT	Yes	OpenNMT(dbrnn) + JPC/Japio corpora + NMT/rule-based posteditor
4	JAPIO	JPCja-en	2017/07/25 18:17:30	1455	-	-	-	0.699930	-	-	-	0.000000	0.000000	0.000000	NMT	No	OpenNMT(dbrnn)
5	ORGANIZER	JPCja-en	2018/08/15 18:38:51	1965	-	-	-	0.699460	-	-	-	-	0.000000	0.000000	NMT	No	NMT with Attention
6	u-tkb	JPCja-en	2017/07/26 12:53:50	1472	-	-	-	0.697290	-	-	-	0.000000	0.000000	0.000000	NMT	No	NMT with SMT phrase translation (phrase extraction with branching entropy; attention over bidirectional LSTMs; by Harvard NMT)
7	bjtu_nlp	JPCja-en	2016/08/16 12:34:36	1149	-	-	-	0.690750	-	-	-	0.000000	0.000000	0.000000	NMT	No	RNN Encoder-Decoder with attention mechanism, single model
8	CUNI	JPCja-en	2017/07/31 22:34:51	1666	-	-	-	0.681520	-	-	-	0.000000	0.000000	0.000000	SMT	No	Bahdanau (2014) seq2seq with conditional GRU on byte-pair encoding
9	ORGANIZER	JPCja-en	2016/07/26 10:15:25	1035	-	-	-	0.673950	-	-	-	0.000000	0.000000	0.000000	Other	Yes	Online A (2016)
10	NICT-2	JPCja-en	2016/08/05 17:58:40	1103	-	-	-	0.672890	-	-	-	0.000000	0.000000	0.000000	SMT	Yes	Phrase-based SMT with Preordering + Domain Adaptation (JPC and ASPEC) + Google 5-gram LM
11	ORGANIZER	JPCja-en	2016/07/13 17:12:09	980	-	-	-	0.672760	-	-	-	0.000000	0.000000	0.000000	SMT	No	String-to-Tree SMT
12	ORGANIZER	JPCja-en	2016/07/13 17:00:31	979	-	-	-	0.672500	-	-	-	0.000000	0.000000	0.000000	SMT	No	Hierarchical Phrase-based SMT
13	Kyoto-U	JPCja-en	2016/07/27 17:15:10	1057	-	-	-	0.672040	-	-	-	0.000000	0.000000	0.000000	EBMT	No	KyotoEBMT 2016 w/o reranking
14	NICT-2	JPCja-en	2016/08/04 17:26:27	1080	-	-	-	0.667540	-	-	-	0.000000	0.000000	0.000000	SMT	No	Phrase-based SMT with Preordering + Domain Adaptation
15	ORGANIZER	JPCja-en	2016/07/13 16:54:09	977	-	-	-	0.664830	-	-	-	0.000000	0.000000	0.000000	SMT	No	Phrase-based SMT
16	Bering Lab	JPCja-en	2021/04/23 13:05:24	5419	-	-	-	0.576681	-	-	-	-	-	-	NMT	Yes	Transformer Ensemble with additional crawled parallel corpus
17	tpt_wat	JPCja-en	2021/04/27 02:32:27	5710	-	-	-	0.569295	-	-	-	-	-	-	NMT	No	Base Transformer model with separate vocab, size 8k
18	ORGANIZER	JPCja-en	2016/08/05 15:18:46	1090	-	-	-	0.521230	-	-	-	0.000000	0.000000	0.000000	Other	Yes	RBMT A (2016)
19	ORGANIZER	JPCja-en	2016/08/05 14:51:27	1088	-	-	-	0.519210	-	-	-	0.000000	0.000000	0.000000	Other	Yes	RBMT C (2016)
20	ORGANIZER	JPCja-en	2016/08/05 15:59:14	1095	-	-	-	0.518110	-	-	-	0.000000	0.000000	0.000000	Other	Yes	RBMT B (2016)
21	ORGANIZER	JPCja-en	2016/07/26 13:43:21	1051	-	-	-	0.486450	-	-	-	0.000000	0.000000	0.000000	Other	Yes	Online B (2016)
22	sarah	JPCja-en	2019/07/26 11:18:48	2969	-	-	-	0.000000	-	-	-	-	-	-	NMT	No	Transformer, ensemble of 4 models
23	KNU_Hyundai	JPCja-en	2019/07/27 12:27:32	3189	-	-	-	0.000000	-	-	-	-	-	-	NMT	Yes	Transformer Base (+ASPEC data), relative position, BT, r2l reranking, checkpoint ensemble
24	goku20	JPCja-en	2020/09/21 12:15:47	4094	-	-	-	0.000000	-	-	-	-	-	-	NMT	No	mBART pre-training transformer, single model
25	goku20	JPCja-en	2020/09/22 00:08:48	4108	-	-	-	0.000000	-	-	-	-	-	-	NMT	No	mBART pre-training transformer, ensemble of 3 models
26	sakura	JPCja-en	2024/08/08 19:55:46	7281	-	-	-	0.000000	-	-	-	-	-	-	NMT	No	LLM: Rakuten/RakutenAI-7B-chat Fine-Tuned with JPC Corpus in six direction (En-Ja, Ja-En, Ko-Ja, Ja-Ko, Zh-Ja, Ja-Zh) - Best
27	sakura	JPCja-en	2024/08/09 00:16:39	7293	-	-	-	0.000000	-	-	-	-	-	-	NMT	No	LLM: Rakuten/RakutenAI-7B-chat Fine-Tuned with JPC Corpus in six direction (En-Ja, Ja-En, Ko-Ja, Ja-Ko, Zh-Ja, Ja-Zh)

Notice:

This table is sorted by the leftmost segmenters. You can change the segmenter used to sort by clicking each segmenter link.
Adequacy-Fluency Metrics (AMFM) is a two-dimensional automatic evaluation metric for machine translation, designed to operate at the sentence level. It is based on adequacy and fluency, to decouple semantic and syntactic components of the translation process to provide a balanced view on translation quality.
AMFM is calculated without tokenizers.
The detail of AMFM is shown on the following paper: "Adequacy–Fluency Metrics: Evaluating MT in the Continuous Space Model Framework" [pdf]. Invited Talk in WAT2015 also helps understanding [slide].

HUMAN (WAT2022)

Notice:

HUMAN (WAT2022) is the result of the Pairwise Crowdsourcing Evaluation on WAT2022.
HUMAN (WAT2022) was evaluated by 5 different workers and the final decision is made by the voting of the judgements.

HUMAN (WAT2021)

Notice:

HUMAN (WAT2021) is the result of the Pairwise Crowdsourcing Evaluation on WAT2021.
HUMAN (WAT2021) was evaluated by 5 different workers and the final decision is made by the voting of the judgements.

HUMAN (WAT2020)

Notice:

HUMAN (WAT2020) is the result of the Pairwise Crowdsourcing Evaluation on WAT2020.
HUMAN (WAT2020) was evaluated by 5 different workers and the final decision is made by the voting of the judgements.

HUMAN (WAT2019)

Notice:

HUMAN (WAT2019) is the result of the Pairwise Crowdsourcing Evaluation on WAT2019.
HUMAN (WAT2019) was evaluated by 5 different workers and the final decision is made by the voting of the judgements.

HUMAN (WAT2018)

Notice:

HUMAN (WAT2018) is the result of the Pairwise Crowdsourcing Evaluation on WAT2018.
HUMAN (WAT2018) was evaluated by 5 different workers and the final decision is made by the voting of the judgements.

HUMAN (WAT2017)

Notice:

HUMAN (WAT2017) is the result of the Pairwise Crowdsourcing Evaluation on WAT2017.
HUMAN (WAT2017) was evaluated by 5 different workers and the final decision is made by the voting of the judgements.

HUMAN (WAT2016)

#	Team	Task	Date/Time	DataID	HUMAN	Method	Other Resources	System Description
1	ORGANIZER	JPCja-en	2016/11/16 11:06:50	1338	71.500	NMT	Yes	Online A (2016/11/14)
2	bjtu_nlp	JPCja-en	2016/08/16 12:34:36	1149	41.500	NMT	No	RNN Encoder-Decoder with attention mechanism, single model
3	ORGANIZER	JPCja-en	2016/07/26 10:15:25	1035	32.250	Other	Yes	Online A (2016)
4	NICT-2	JPCja-en	2016/08/04 17:26:27	1080	25.000	SMT	No	Phrase-based SMT with Preordering + Domain Adaptation
5	NICT-2	JPCja-en	2016/08/05 17:58:40	1103	24.250	SMT	Yes	Phrase-based SMT with Preordering + Domain Adaptation (JPC and ASPEC) + Google 5-gram LM
6	ORGANIZER	JPCja-en	2016/08/05 15:18:46	1090	23.750	Other	Yes	RBMT A (2016)
7	ORGANIZER	JPCja-en	2016/07/13 17:12:09	980	23.000	SMT	No	String-to-Tree SMT
8	ORGANIZER	JPCja-en	2016/07/13 17:00:31	979	8.750	SMT	No	Hierarchical Phrase-based SMT

Notice:

HUMAN (WAT2016) is the result of the Pairwise Crowdsourcing Evaluation on WAT2016.
HUMAN (WAT2016) was evaluated by 5 different workers and the final decision is made by the voting of the judgements.

HUMAN (WAT2015)

Notice:

HUMAN (WAT2015) is the result of the Pairwise Crowdsourcing Evaluation on WAT2015.
HUMAN (WAT2015) was evaluated by 5 different workers and the final decision is made by the voting of the judgements.
The detail of the evaluation can be found in the pdf document (PDF file).

HUMAN (WAT2014)

Notice:

HUMAN (WAT2014) is the result of the Pairwise Crowdsourcing Evaluation on WAT2014.
HUMAN (WAT2014) was evaluated by 3 different workers and the final decision is made by the voting of the judgements.
The detail of the evaluation can be found in the pdf document (PDF file).

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

WAT The Workshop on Asian Translation Evaluation Results

BLEU

RIBES

AMFM

HUMAN (WAT2022)

HUMAN (WAT2021)

HUMAN (WAT2020)

HUMAN (WAT2019)

HUMAN (WAT2018)

HUMAN (WAT2017)

HUMAN (WAT2016)

HUMAN (WAT2015)

HUMAN (WAT2014)

EVALUATION RESULTS USAGE POLICY

WAT

The Workshop on Asian Translation

Evaluation Results