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	nictrb	ASPECRTen-ja	2021/04/20 15:31:08	5216	48.92	49.24	49.25	-	-	-	-	-	-	-	NMT	Yes	Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
2	NTT	ASPECRTen-ja	2021/04/21 16:24:55	5296	53.47	54.20	53.79	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer + data aug + ptrnet + beam search, Single model
3	NHK	ASPECRTen-ja	2021/04/21 17:33:15	5303	44.56	46.49	45.25	-	-	-	-	-	-	-	NMT	No	baseline (using aspec3) w/o constraint
4	NHK	ASPECRTen-ja	2021/04/21 17:43:15	5305	43.90	45.93	44.64	-	-	-	-	-	-	-	NMT	No	baseline w/o constraint
5	NTT	ASPECRTen-ja	2021/04/21 18:16:10	5313	55.26	55.81	55.49	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer + data aug + ptrnet + constrained decoding, Single model
6	NTT	ASPECRTen-ja	2021/04/22 12:33:16	5368	56.39	56.87	56.57	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer + data aug + ptrnet + constrained decoding, 8 ensemble
7	NHK	ASPECRTen-ja	2021/04/25 00:22:24	5502	52.07	52.69	52.33	-	-	-	-	-	-	-	NMT	No	lexical-constrained with filter + add all list with rule
8	nictrb	ASPECRTen-ja	2021/04/26 00:23:43	5589	50.93	51.18	51.21	-	-	-	-	-	-	-	NMT	Yes	Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
9	nictrb	ASPECRTen-ja	2021/04/26 01:22:48	5591	51.07	51.32	51.36	-	-	-	-	-	-	-	NMT	Yes	Vanilla+Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
10	NTT	ASPECRTen-ja	2021/04/26 14:32:16	5616	56.20	56.67	56.47	-	-	-	-	-	-	-	NMT	Yes	ASPEC first 2M + BT 1M + JParaCrawl + CC, Transformer + data aug + ptrnet + constrained decoding + fine-tune, 8 ensemble
11	NTT	ASPECRTen-ja	2021/04/26 20:57:47	5664	43.92	45.91	44.68	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer big, Single model
12	NTT	ASPECRTen-ja	2021/04/26 21:00:40	5665	45.08	45.68	45.38	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer big + constrained decoding, Single model
13	NHK	ASPECRTen-ja	2021/05/23 00:51:09	6404	36.33	37.43	36.53	-	-	-	-	-	-	-	NMT	No	Random-word constraint
14	TMU	ASPECRTen-ja	2022/07/11 22:22:50	6798	49.40	49.40	49.60	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + levT
15	TMU	ASPECRTen-ja	2022/07/16 19:49:05	6946	49.70	49.60	49.90	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (distillation pretrain)
16	TMU	ASPECRTen-ja	2022/07/16 19:51:15	6947	49.80	49.80	50.00	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (mix)
17	TMU	ASPECRTen-ja	2022/07/16 19:54:27	6948	50.80	51.60	51.30	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet
18	TMU	ASPECRTen-ja	2022/07/16 20:59:14	6953	52.00	52.10	52.20	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + 5model ensemble + LevT (mix)

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	nictrb	ASPECRTen-ja	2021/04/20 15:31:08	5216	0.827925	0.831527	0.835757	-	-	-	-	-	-	-	NMT	Yes	Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
2	NTT	ASPECRTen-ja	2021/04/21 16:24:55	5296	0.877759	0.878210	0.882739	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer + data aug + ptrnet + beam search, Single model
3	NHK	ASPECRTen-ja	2021/04/21 17:33:15	5303	0.856942	0.857427	0.863199	-	-	-	-	-	-	-	NMT	No	baseline (using aspec3) w/o constraint
4	NHK	ASPECRTen-ja	2021/04/21 17:43:15	5305	0.852867	0.854444	0.859958	-	-	-	-	-	-	-	NMT	No	baseline w/o constraint
5	NTT	ASPECRTen-ja	2021/04/21 18:16:10	5313	0.875952	0.876317	0.880940	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer + data aug + ptrnet + constrained decoding, Single model
6	NTT	ASPECRTen-ja	2021/04/22 12:33:16	5368	0.882454	0.882322	0.887104	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer + data aug + ptrnet + constrained decoding, 8 ensemble
7	NHK	ASPECRTen-ja	2021/04/25 00:22:24	5502	0.815612	0.823084	0.827300	-	-	-	-	-	-	-	NMT	No	lexical-constrained with filter + add all list with rule
8	nictrb	ASPECRTen-ja	2021/04/26 00:23:43	5589	0.832776	0.835202	0.840048	-	-	-	-	-	-	-	NMT	Yes	Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
9	nictrb	ASPECRTen-ja	2021/04/26 01:22:48	5591	0.836874	0.839934	0.844141	-	-	-	-	-	-	-	NMT	Yes	Vanilla+Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
10	NTT	ASPECRTen-ja	2021/04/26 14:32:16	5616	0.885308	0.885612	0.889831	-	-	-	-	-	-	-	NMT	Yes	ASPEC first 2M + BT 1M + JParaCrawl + CC, Transformer + data aug + ptrnet + constrained decoding + fine-tune, 8 ensemble
11	NTT	ASPECRTen-ja	2021/04/26 20:57:47	5664	0.852597	0.856560	0.859844	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer big, Single model
12	NTT	ASPECRTen-ja	2021/04/26 21:00:40	5665	0.840177	0.841064	0.845364	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer big + constrained decoding, Single model
13	NHK	ASPECRTen-ja	2021/05/23 00:51:09	6404	0.794851	0.802395	0.807206	-	-	-	-	-	-	-	NMT	No	Random-word constraint
14	TMU	ASPECRTen-ja	2022/07/11 22:22:50	6798	0.863900	0.864670	0.868942	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + levT
15	TMU	ASPECRTen-ja	2022/07/16 19:49:05	6946	0.864646	0.865736	0.869766	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (distillation pretrain)
16	TMU	ASPECRTen-ja	2022/07/16 19:51:15	6947	0.864394	0.865434	0.869480	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (mix)
17	TMU	ASPECRTen-ja	2022/07/16 19:54:27	6948	0.867732	0.869859	0.873110	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet
18	TMU	ASPECRTen-ja	2022/07/16 20:59:14	6953	0.872152	0.872446	0.877425	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + 5model ensemble + LevT (mix)

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	nictrb	ASPECRTen-ja	2021/04/20 15:31:08	5216	0.791510	0.791510	0.791510	-	-	-	-	-	-	-	NMT	Yes	Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
2	NTT	ASPECRTen-ja	2021/04/21 16:24:55	5296	0.811200	0.811200	0.811200	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer + data aug + ptrnet + beam search, Single model
3	NHK	ASPECRTen-ja	2021/04/21 17:33:15	5303	0.760880	0.760880	0.760880	-	-	-	-	-	-	-	NMT	No	baseline (using aspec3) w/o constraint
4	NHK	ASPECRTen-ja	2021/04/21 17:43:15	5305	0.756770	0.756770	0.756770	-	-	-	-	-	-	-	NMT	No	baseline w/o constraint
5	NTT	ASPECRTen-ja	2021/04/21 18:16:10	5313	0.813970	0.813970	0.813970	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer + data aug + ptrnet + constrained decoding, Single model
6	NTT	ASPECRTen-ja	2021/04/22 12:33:16	5368	0.817290	0.817290	0.817290	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer + data aug + ptrnet + constrained decoding, 8 ensemble
7	NHK	ASPECRTen-ja	2021/04/25 00:22:24	5502	0.801660	0.801660	0.801660	-	-	-	-	-	-	-	NMT	No	lexical-constrained with filter + add all list with rule
8	nictrb	ASPECRTen-ja	2021/04/26 00:23:43	5589	0.799120	0.799120	0.799120	-	-	-	-	-	-	-	NMT	Yes	Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
9	nictrb	ASPECRTen-ja	2021/04/26 01:22:48	5591	0.799950	0.799950	0.799950	-	-	-	-	-	-	-	NMT	Yes	Vanilla+Sparse Transformer-big+Sampling as Focus Training+Constrained Decoding (w/ ParaCrawl-v5.1 + News Commentary v15 for NMT pre-training)
10	NTT	ASPECRTen-ja	2021/04/26 14:32:16	5616	0.818190	0.818190	0.818190	-	-	-	-	-	-	-	NMT	Yes	ASPEC first 2M + BT 1M + JParaCrawl + CC, Transformer + data aug + ptrnet + constrained decoding + fine-tune, 8 ensemble
11	NTT	ASPECRTen-ja	2021/04/26 20:57:47	5664	0.765480	0.765480	0.765480	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer big, Single model
12	NTT	ASPECRTen-ja	2021/04/26 21:00:40	5665	0.798960	0.798960	0.798960	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M + BT 1M, Transformer big + constrained decoding, Single model
13	NHK	ASPECRTen-ja	2021/05/23 00:51:09	6404	0.700610	0.700610	0.700610	-	-	-	-	-	-	-	NMT	No	Random-word constraint
14	TMU	ASPECRTen-ja	2022/07/11 22:22:50	6798	0.785920	0.785920	0.785920	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + levT
15	TMU	ASPECRTen-ja	2022/07/16 19:49:05	6946	0.788770	0.788770	0.788770	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (distillation pretrain)
16	TMU	ASPECRTen-ja	2022/07/16 19:51:15	6947	0.788640	0.788640	0.788640	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + LevT (mix)
17	TMU	ASPECRTen-ja	2022/07/16 19:54:27	6948	0.800450	0.800450	0.800450	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet
18	TMU	ASPECRTen-ja	2022/07/16 20:59:14	6953	0.800580	0.800580	0.800580	-	-	-	-	-	-	-	NMT	No	ASPEC first 2M, Transformer + data aug + ptrnet + 5model ensemble + LevT (mix)

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)

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