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	NTT	ASPECRTja-en	2021/04/20 13:52:42	5209	-	-	-	44.34	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + constrained decoding, 8 ensemble
2	NTT	ASPECRTja-en	2021/04/26 14:30:46	5615	-	-	-	44.28	-	-	-	-	-	-	NMT	Yes	ASPEC first 2M + FW 1M + JParaCrawl + CC, Transformer + data aug + ptrnet + constrained decoding + fine-tune, 8 ensemble
3	NTT	ASPECRTja-en	2021/04/21 15:54:44	5294	-	-	-	44.26	-	-	-	-	-	-	NMT	No	ASPEC first 2M + FW 1M, Transformer + data aug + ptrnet + constrained decoding, 8 ensemble
4	NTT	ASPECRTja-en	2021/04/21 18:10:58	5312	-	-	-	43.33	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + constrained decoding, Single model
5	NHK	ASPECRTja-en	2021/04/25 00:27:35	5505	-	-	-	42.94	-	-	-	-	-	-	NMT	No	lexical-constrained with filter + add all list with rule
6	TMU	ASPECRTja-en	2022/07/16 18:04:57	6945	-	-	-	42.30	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + 5model ensemble + LevT (mix)
7	NTT	ASPECRTja-en	2021/04/21 16:16:14	5295	-	-	-	41.88	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet, Single model
8	TMU	ASPECRTja-en	2022/07/16 17:50:07	6941	-	-	-	39.60	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (distillation pretrain)
9	TMU	ASPECRTja-en	2022/07/16 17:56:25	6942	-	-	-	39.60	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (mix)
10	TMU	ASPECRTja-en	2022/07/22 11:15:29	6981	-	-	-	39.50	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + levT
11	TMU	ASPECRTja-en	2022/07/16 19:56:31	6949	-	-	-	39.30	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet
12	nictrb	ASPECRTja-en	2021/04/26 02:10:46	5592	-	-	-	37.01	-	-	-	-	-	-	NMT	Yes	Vanilla+Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
13	nictrb	ASPECRTja-en	2021/04/26 01:09:28	5590	-	-	-	36.83	-	-	-	-	-	-	NMT	Yes	Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
14	nictrb	ASPECRTja-en	2021/04/20 16:27:31	5220	-	-	-	35.26	-	-	-	-	-	-	NMT	Yes	Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
15	NHK	ASPECRTja-en	2021/04/21 17:50:29	5309	-	-	-	29.61	-	-	-	-	-	-	NMT	No	baseline (using aspec3) w/o constraint
16	NHK	ASPECRTja-en	2021/05/23 00:53:53	6405	-	-	-	29.49	-	-	-	-	-	-	NMT	No	Random-word constraint
17	NTT	ASPECRTja-en	2021/04/26 21:02:26	5668	-	-	-	29.30	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer big w/o constrained decoding, Single model
18	NHK	ASPECRTja-en	2021/04/21 17:52:46	5311	-	-	-	29.25	-	-	-	-	-	-	NMT	No	baseline w/o constraint
19	TMU	ASPECRTja-en	2021/05/03 13:23:49	5994	-	-	-	25.29	-	-	-	-	-	-	NMT	No	RecoverSAT (seg:14) + Force Translation + RTV ordering (w/ Bitext 1.5M + BT 1.5M)
20	NTT	ASPECRTja-en	2021/04/26 21:04:38	5670	-	-	-	23.22	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer big + constrained decoding, Single model

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	NTT	ASPECRTja-en	2021/04/21 15:54:44	5294	-	-	-	0.813989	-	-	-	-	-	-	NMT	No	ASPEC first 2M + FW 1M, Transformer + data aug + ptrnet + constrained decoding, 8 ensemble
2	NTT	ASPECRTja-en	2021/04/26 14:30:46	5615	-	-	-	0.813155	-	-	-	-	-	-	NMT	Yes	ASPEC first 2M + FW 1M + JParaCrawl + CC, Transformer + data aug + ptrnet + constrained decoding + fine-tune, 8 ensemble
3	NTT	ASPECRTja-en	2021/04/20 13:52:42	5209	-	-	-	0.811700	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + constrained decoding, 8 ensemble
4	NTT	ASPECRTja-en	2021/04/21 16:16:14	5295	-	-	-	0.811303	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet, Single model
5	NTT	ASPECRTja-en	2021/04/21 18:10:58	5312	-	-	-	0.810012	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + constrained decoding, Single model
6	TMU	ASPECRTja-en	2022/07/16 18:04:57	6945	-	-	-	0.808360	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + 5model ensemble + LevT (mix)
7	NHK	ASPECRTja-en	2021/04/25 00:27:35	5505	-	-	-	0.801015	-	-	-	-	-	-	NMT	No	lexical-constrained with filter + add all list with rule
8	TMU	ASPECRTja-en	2022/07/16 17:50:07	6941	-	-	-	0.799612	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (distillation pretrain)
9	TMU	ASPECRTja-en	2022/07/22 11:15:29	6981	-	-	-	0.799591	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + levT
10	TMU	ASPECRTja-en	2022/07/16 17:56:25	6942	-	-	-	0.799376	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (mix)
11	TMU	ASPECRTja-en	2022/07/16 19:56:31	6949	-	-	-	0.795787	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet
12	NHK	ASPECRTja-en	2021/04/21 17:50:29	5309	-	-	-	0.771441	-	-	-	-	-	-	NMT	No	baseline (using aspec3) w/o constraint
13	NHK	ASPECRTja-en	2021/04/21 17:52:46	5311	-	-	-	0.770681	-	-	-	-	-	-	NMT	No	baseline w/o constraint
14	NTT	ASPECRTja-en	2021/04/26 21:02:26	5668	-	-	-	0.770525	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer big w/o constrained decoding, Single model
15	nictrb	ASPECRTja-en	2021/04/26 01:09:28	5590	-	-	-	0.758458	-	-	-	-	-	-	NMT	Yes	Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
16	nictrb	ASPECRTja-en	2021/04/26 02:10:46	5592	-	-	-	0.753823	-	-	-	-	-	-	NMT	Yes	Vanilla+Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
17	nictrb	ASPECRTja-en	2021/04/20 16:27:31	5220	-	-	-	0.744416	-	-	-	-	-	-	NMT	Yes	Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
18	NTT	ASPECRTja-en	2021/04/26 21:04:38	5670	-	-	-	0.719109	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer big + constrained decoding, Single model
19	NHK	ASPECRTja-en	2021/05/23 00:53:53	6405	-	-	-	0.683448	-	-	-	-	-	-	NMT	No	Random-word constraint
20	TMU	ASPECRTja-en	2021/05/03 13:23:49	5994	-	-	-	0.653597	-	-	-	-	-	-	NMT	No	RecoverSAT (seg:14) + Force Translation + RTV ordering (w/ Bitext 1.5M + BT 1.5M)

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	NTT	ASPECRTja-en	2021/04/21 15:54:44	5294	-	-	-	0.676740	-	-	-	-	-	-	NMT	No	ASPEC first 2M + FW 1M, Transformer + data aug + ptrnet + constrained decoding, 8 ensemble
2	NTT	ASPECRTja-en	2021/04/26 14:30:46	5615	-	-	-	0.676670	-	-	-	-	-	-	NMT	Yes	ASPEC first 2M + FW 1M + JParaCrawl + CC, Transformer + data aug + ptrnet + constrained decoding + fine-tune, 8 ensemble
3	NTT	ASPECRTja-en	2021/04/20 13:52:42	5209	-	-	-	0.672320	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + constrained decoding, 8 ensemble
4	NTT	ASPECRTja-en	2021/04/21 18:10:58	5312	-	-	-	0.671200	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + constrained decoding, Single model
5	NTT	ASPECRTja-en	2021/04/21 16:16:14	5295	-	-	-	0.670140	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet, Single model
6	NHK	ASPECRTja-en	2021/04/25 00:27:35	5505	-	-	-	0.661560	-	-	-	-	-	-	NMT	No	lexical-constrained with filter + add all list with rule
7	TMU	ASPECRTja-en	2022/07/16 18:04:57	6945	-	-	-	0.653910	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + 5model ensemble + LevT (mix)
8	TMU	ASPECRTja-en	2022/07/16 19:56:31	6949	-	-	-	0.653280	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet
9	nictrb	ASPECRTja-en	2021/04/26 01:09:28	5590	-	-	-	0.652980	-	-	-	-	-	-	NMT	Yes	Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
10	nictrb	ASPECRTja-en	2021/04/26 02:10:46	5592	-	-	-	0.651570	-	-	-	-	-	-	NMT	Yes	Vanilla+Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
11	nictrb	ASPECRTja-en	2021/04/20 16:27:31	5220	-	-	-	0.641620	-	-	-	-	-	-	NMT	Yes	Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
12	TMU	ASPECRTja-en	2022/07/22 11:15:29	6981	-	-	-	0.641300	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + levT
13	TMU	ASPECRTja-en	2022/07/16 17:56:25	6942	-	-	-	0.640000	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (mix)
14	TMU	ASPECRTja-en	2022/07/16 17:50:07	6941	-	-	-	0.638150	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (distillation pretrain)
15	NTT	ASPECRTja-en	2021/04/26 21:04:38	5670	-	-	-	0.636660	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer big + constrained decoding, Single model
16	NHK	ASPECRTja-en	2021/04/21 17:50:29	5309	-	-	-	0.623370	-	-	-	-	-	-	NMT	No	baseline (using aspec3) w/o constraint
17	NTT	ASPECRTja-en	2021/04/26 21:02:26	5668	-	-	-	0.619200	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer big w/o constrained decoding, Single model
18	NHK	ASPECRTja-en	2021/04/21 17:52:46	5311	-	-	-	0.617700	-	-	-	-	-	-	NMT	No	baseline w/o constraint
19	TMU	ASPECRTja-en	2021/05/03 13:23:49	5994	-	-	-	0.612290	-	-	-	-	-	-	NMT	No	RecoverSAT (seg:14) + Force Translation + RTV ordering (w/ Bitext 1.5M + BT 1.5M)
20	NHK	ASPECRTja-en	2021/05/23 00:53:53	6405	-	-	-	0.518210	-	-	-	-	-	-	NMT	No	Random-word constraint

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)

#	Team	Task	Date/Time	DataID	HUMAN	Method	Other Resources	System Description
1	NTT	ASPECRTja-en	2021/04/20 13:52:42	5209	Underway	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + constrained decoding, 8 ensemble
2	NHK	ASPECRTja-en	2021/04/25 00:27:35	5505	Underway	NMT	No	lexical-constrained with filter + add all list with rule
3	nictrb	ASPECRTja-en	2021/04/26 02:10:46	5592	Underway	NMT	Yes	Vanilla+Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
4	NTT	ASPECRTja-en	2021/04/26 14:30:46	5615	Underway	NMT	Yes	ASPEC first 2M + FW 1M + JParaCrawl + CC, Transformer + data aug + ptrnet + constrained decoding + fine-tune, 8 ensemble
5	TMU	ASPECRTja-en	2021/05/03 13:23:49	5994	Underway	NMT	No	RecoverSAT (seg:14) + Force Translation + RTV ordering (w/ Bitext 1.5M + BT 1.5M)

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)

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