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	NICT-4	ALTen-my	2018/09/13 14:51:40	2294	-	-	-	-	-	-	-	32.30	0.00	0.00	Other	Yes	Many PBSMT and NMT n-best lists combined and reranked. Use noisy Wikipedia data for back-translation and language model trainings.
2	NICT-4	ALTen-my	2018/09/13 14:20:19	2287	-	-	-	-	-	-	-	30.52	0.00	0.00	Other	No	Many PBSMT and NMT n-best lists combined and reranked
3	NICT	ALTen-my	2018/09/14 18:07:14	2345	-	-	-	-	-	-	-	29.89	0.00	0.00	NMT	No	4 model ensemble
4	NICT-4	ALTen-my	2018/08/24 11:07:36	2087	-	-	-	-	-	-	-	29.57	0.00	0.00	NMT	No	Ensemble of 4 models
5	NICT-4	ALTen-my	2018/08/24 10:36:46	2084	-	-	-	-	-	-	-	27.61	0.00	0.00	NMT	No	single model
6	NICT	ALTen-my	2018/09/12 15:45:56	2282	-	-	-	-	-	-	-	26.02	0.00	0.00	NMT	No	Single model
7	NICT-4	ALTen-my	2018/09/13 14:24:47	2288	-	-	-	-	-	-	-	23.14	0.00	0.00	SMT	No	with MSLR models, language models were trained on the target side of the parallel data
8	UCSYNLP	ALTen-my	2018/09/14 17:18:59	2341	-	-	-	-	-	-	-	22.78	0.00	0.00	SMT	No	OSM
9	XMUNLP	ALTen-my	2018/09/15 16:40:17	2398	-	-	-	-	-	-	-	22.76	0.00	0.00	NMT	No	single rnnsearch model
10	ORGANIZER	ALTen-my	2018/09/04 18:37:12	2227	-	-	-	-	-	-	-	22.42	0.00	0.00	NMT	No	NMT with Attention
11	UCSYNLP	ALTen-my	2018/09/14 17:52:12	2344	-	-	-	-	-	-	-	22.40	0.00	0.00	SMT	No	PBSMT
12	Osaka-U	ALTen-my	2018/09/15 22:57:16	2437	-	-	-	-	-	-	-	22.33	0.00	0.00	NMT	Yes	rewarding model
13	XMUNLP	ALTen-my	2018/09/16 08:45:01	2455	-	-	-	-	-	-	-	21.57	0.00	0.00	NMT	No	single transformer model
14	UCSYNLP	ALTen-my	2018/09/14 15:54:51	2339	-	-	-	-	-	-	-	21.19	0.00	0.00	NMT	No	Transformer
15	Osaka-U	ALTen-my	2018/09/16 13:08:36	2471	-	-	-	-	-	-	-	20.88	0.00	0.00	SMT	No	preordering with neural network
16	ORGANIZER	ALTen-my	2018/08/24 15:32:01	2143	-	-	-	-	-	-	-	20.83	0.00	0.00	Other	Yes	Online A (comma -> 0x104a)
17	ORGANIZER	ALTen-my	2018/08/24 15:31:02	2142	-	-	-	-	-	-	-	20.31	0.00	0.00	Other	Yes	Online A
18	kmust88	ALTen-my	2018/09/15 00:12:28	2360	-	-	-	-	-	-	-	19.34	0.00	0.00	NMT	No	training the model base on transformer and do some
19	UCSYNLP	ALTen-my	2018/09/14 15:56:00	2340	-	-	-	-	-	-	-	19.19	0.00	0.00	NMT	No	NMT with Attention
20	UCSMNLP	ALTen-my	2018/10/29 15:32:37	2550	-	-	-	-	-	-	-	19.17	0.00	0.00	SMT	No	Batch MIRA tuning
21	Osaka-U	ALTen-my	2018/09/16 11:55:07	2462	-	-	-	-	-	-	-	9.45	0.00	0.00	NMT	No	mixed fine tuning
22	UCSMNLP	ALTen-my	2018/09/14 15:27:26	2337	-	-	-	-	-	-	-	8.16	0.00	0.00	SMT	No	with PBSMT

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	NICT-4	ALTen-my	2018/09/13 14:51:40	2294	-	-	-	-	-	-	-	0.746480	0.000000	0.000000	Other	Yes	Many PBSMT and NMT n-best lists combined and reranked. Use noisy Wikipedia data for back-translation and language model trainings.
2	NICT-4	ALTen-my	2018/08/24 11:07:36	2087	-	-	-	-	-	-	-	0.738538	0.000000	0.000000	NMT	No	Ensemble of 4 models
3	NICT-4	ALTen-my	2018/09/13 14:20:19	2287	-	-	-	-	-	-	-	0.733501	0.000000	0.000000	Other	No	Many PBSMT and NMT n-best lists combined and reranked
4	NICT	ALTen-my	2018/09/14 18:07:14	2345	-	-	-	-	-	-	-	0.726922	0.000000	0.000000	NMT	No	4 model ensemble
5	NICT-4	ALTen-my	2018/08/24 10:36:46	2084	-	-	-	-	-	-	-	0.725103	0.000000	0.000000	NMT	No	single model
6	NICT	ALTen-my	2018/09/12 15:45:56	2282	-	-	-	-	-	-	-	0.694652	0.000000	0.000000	NMT	No	Single model
7	ORGANIZER	ALTen-my	2018/08/24 15:32:01	2143	-	-	-	-	-	-	-	0.679968	0.000000	0.000000	Other	Yes	Online A (comma -> 0x104a)
8	UCSYNLP	ALTen-my	2018/09/14 15:54:51	2339	-	-	-	-	-	-	-	0.679800	0.000000	0.000000	NMT	No	Transformer
9	ORGANIZER	ALTen-my	2018/08/24 15:31:02	2142	-	-	-	-	-	-	-	0.678360	0.000000	0.000000	Other	Yes	Online A
10	XMUNLP	ALTen-my	2018/09/15 16:40:17	2398	-	-	-	-	-	-	-	0.674455	0.000000	0.000000	NMT	No	single rnnsearch model
11	UCSYNLP	ALTen-my	2018/09/14 15:56:00	2340	-	-	-	-	-	-	-	0.671461	0.000000	0.000000	NMT	No	NMT with Attention
12	Osaka-U	ALTen-my	2018/09/15 22:57:16	2437	-	-	-	-	-	-	-	0.668596	0.000000	0.000000	NMT	Yes	rewarding model
13	XMUNLP	ALTen-my	2018/09/16 08:45:01	2455	-	-	-	-	-	-	-	0.668210	0.000000	0.000000	NMT	No	single transformer model
14	ORGANIZER	ALTen-my	2018/09/04 18:37:12	2227	-	-	-	-	-	-	-	0.667437	0.000000	0.000000	NMT	No	NMT with Attention
15	kmust88	ALTen-my	2018/09/15 00:12:28	2360	-	-	-	-	-	-	-	0.650796	0.000000	0.000000	NMT	No	training the model base on transformer and do some
16	Osaka-U	ALTen-my	2018/09/16 13:08:36	2471	-	-	-	-	-	-	-	0.639517	0.000000	0.000000	SMT	No	preordering with neural network
17	Osaka-U	ALTen-my	2018/09/16 11:55:07	2462	-	-	-	-	-	-	-	0.581903	0.000000	0.000000	NMT	No	mixed fine tuning
18	NICT-4	ALTen-my	2018/09/13 14:24:47	2288	-	-	-	-	-	-	-	0.565243	0.000000	0.000000	SMT	No	with MSLR models, language models were trained on the target side of the parallel data
19	UCSMNLP	ALTen-my	2018/10/29 15:32:37	2550	-	-	-	-	-	-	-	0.554023	0.000000	0.000000	SMT	No	Batch MIRA tuning
20	UCSYNLP	ALTen-my	2018/09/14 17:18:59	2341	-	-	-	-	-	-	-	0.549883	0.000000	0.000000	SMT	No	OSM
21	UCSYNLP	ALTen-my	2018/09/14 17:52:12	2344	-	-	-	-	-	-	-	0.544395	0.000000	0.000000	SMT	No	PBSMT
22	UCSMNLP	ALTen-my	2018/09/14 15:27:26	2337	-	-	-	-	-	-	-	0.470758	0.000000	0.000000	SMT	No	with PBSMT

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	NICT-4	ALTen-my	2018/09/13 14:51:40	2294	-	-	-	-	-	-	-	0.816480	0.000000	0.000000	Other	Yes	Many PBSMT and NMT n-best lists combined and reranked. Use noisy Wikipedia data for back-translation and language model trainings.
2	NICT-4	ALTen-my	2018/09/13 14:20:19	2287	-	-	-	-	-	-	-	0.809750	0.000000	0.000000	Other	No	Many PBSMT and NMT n-best lists combined and reranked
3	NICT-4	ALTen-my	2018/08/24 11:07:36	2087	-	-	-	-	-	-	-	0.803810	0.000000	0.000000	NMT	No	Ensemble of 4 models
4	NICT	ALTen-my	2018/09/14 18:07:14	2345	-	-	-	-	-	-	-	0.800230	0.000000	0.000000	NMT	No	4 model ensemble
5	NICT-4	ALTen-my	2018/08/24 10:36:46	2084	-	-	-	-	-	-	-	0.799000	0.000000	0.000000	NMT	No	single model
6	NICT	ALTen-my	2018/09/12 15:45:56	2282	-	-	-	-	-	-	-	0.785920	0.000000	0.000000	NMT	No	Single model
7	Osaka-U	ALTen-my	2018/09/16 13:08:36	2471	-	-	-	-	-	-	-	0.774750	0.000000	0.000000	SMT	No	preordering with neural network
8	XMUNLP	ALTen-my	2018/09/16 08:45:01	2455	-	-	-	-	-	-	-	0.772120	0.000000	0.000000	NMT	No	single transformer model
9	UCSYNLP	ALTen-my	2018/09/14 15:54:51	2339	-	-	-	-	-	-	-	0.756710	0.000000	0.000000	NMT	No	Transformer
10	UCSYNLP	ALTen-my	2018/09/14 17:18:59	2341	-	-	-	-	-	-	-	0.751180	0.000000	0.000000	SMT	No	OSM
11	UCSYNLP	ALTen-my	2018/09/14 17:52:12	2344	-	-	-	-	-	-	-	0.749080	0.000000	0.000000	SMT	No	PBSMT
12	XMUNLP	ALTen-my	2018/09/15 16:40:17	2398	-	-	-	-	-	-	-	0.748940	0.000000	0.000000	NMT	No	single rnnsearch model
13	ORGANIZER	ALTen-my	2018/09/04 18:37:12	2227	-	-	-	-	-	-	-	0.745550	0.000000	0.000000	NMT	No	NMT with Attention
14	Osaka-U	ALTen-my	2018/09/15 22:57:16	2437	-	-	-	-	-	-	-	0.740760	0.000000	0.000000	NMT	Yes	rewarding model
15	kmust88	ALTen-my	2018/09/15 00:12:28	2360	-	-	-	-	-	-	-	0.721280	0.000000	0.000000	NMT	No	training the model base on transformer and do some
16	UCSYNLP	ALTen-my	2018/09/14 15:56:00	2340	-	-	-	-	-	-	-	0.717480	0.000000	0.000000	NMT	No	NMT with Attention
17	Osaka-U	ALTen-my	2018/09/16 11:55:07	2462	-	-	-	-	-	-	-	0.665360	0.000000	0.000000	NMT	No	mixed fine tuning
18	NICT-4	ALTen-my	2018/09/13 14:24:47	2288	-	-	-	-	-	-	-	0.618380	0.000000	0.000000	SMT	No	with MSLR models, language models were trained on the target side of the parallel data
19	UCSMNLP	ALTen-my	2018/10/29 15:32:37	2550	-	-	-	-	-	-	-	0.615240	0.000000	0.000000	SMT	No	Batch MIRA tuning
20	ORGANIZER	ALTen-my	2018/08/24 15:32:01	2143	-	-	-	-	-	-	-	0.594230	0.000000	0.000000	Other	Yes	Online A (comma -> 0x104a)
21	ORGANIZER	ALTen-my	2018/08/24 15:31:02	2142	-	-	-	-	-	-	-	0.587120	0.000000	0.000000	Other	Yes	Online A
22	UCSMNLP	ALTen-my	2018/09/14 15:27:26	2337	-	-	-	-	-	-	-	0.222510	0.000000	0.000000	SMT	No	with PBSMT

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)

#	Team	Task	Date/Time	DataID	HUMAN	Method	Other Resources	System Description
1	NICT	ALTen-my	2018/09/14 18:07:14	2345	61.000	NMT	No	4 model ensemble
2	NICT-4	ALTen-my	2018/08/24 11:07:36	2087	53.000	NMT	No	Ensemble of 4 models
3	NICT	ALTen-my	2018/09/12 15:45:56	2282	42.500	NMT	No	Single model
4	NICT-4	ALTen-my	2018/09/13 14:20:19	2287	39.750	Other	No	Many PBSMT and NMT n-best lists combined and reranked
5	UCSYNLP	ALTen-my	2018/09/14 15:54:51	2339	10.500	NMT	No	Transformer
6	kmust88	ALTen-my	2018/09/15 00:12:28	2360	9.750	NMT	No	training the model base on transformer and do some
7	Osaka-U	ALTen-my	2018/09/15 22:57:16	2437	3.000	NMT	Yes	rewarding model
8	UCSYNLP	ALTen-my	2018/09/14 15:56:00	2340	0.750	NMT	No	NMT with Attention
9	Osaka-U	ALTen-my	2018/09/16 13:08:36	2471	-23.500	SMT	No	preordering with neural network
10	UCSMNLP	ALTen-my	2018/09/14 15:27:26	2337	-96.750	SMT	No	with PBSMT

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