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	TMEKU	MMTen-ja	2021/04/01 09:43:23	4730	44.95	53.50	48.57	-	-	-	-	-	-	-	NMT	No	RNN-based MNMT model. Ensemble top 10 models.
2	TMEKU	MMTen-ja	2021/04/23 22:24:24	5452	43.40	51.81	47.02	-	-	-	-	-	-	-	NMT	No	RNN-based MNMT model. Single model.
3	sakura	MMTen-ja	2021/05/04 18:29:15	6313	43.09	51.17	46.32	-	-	-	-	-	-	-	NMT	No	The ensemble of two transformer-based MNMT models with universal visual representation
4	sakura	MMTen-ja	2021/05/04 15:58:06	6256	42.92	50.79	46.02	-	-	-	-	-	-	-	NMT	Yes	Text-only mBART25 finetuning
5	HW-TSC	MMTen-ja	2020/09/18 17:08:15	3914	41.26	49.79	44.76	-	-	-	-	-	-	-	NMT	Yes	mmt_cmask_pretrained_transformer_big
6	TMU	MMTen-ja	2020/09/08 17:24:15	3651	40.71	49.81	44.57	-	-	-	-	-	-	-	NMT	No	The Ensemble of top6 models.
7	TMU	MMTen-ja	2020/09/15 15:21:42	3743	40.16	49.23	44.12	-	-	-	-	-	-	-	NMT	No	Multimodal NMT (BiGRU + ResNet50; decinit) pretrained on noised F30kEnt-JP training data, fine-tuned on clean data.
8	TMU	MMTen-ja	2020/09/08 17:29:27	3653	40.02	49.31	43.99	-	-	-	-	-	-	-	NMT	No	The Ensemble of 3 baselines. Baseline: Multimodal NMT (BiGRU + ResNet50; decinit)
9	TMU	MMTen-ja	2020/09/19 11:54:51	4046	39.92	48.94	43.78	-	-	-	-	-	-	-	NMT	No	The Ensemble of 3 models. model: Multimodal NMT (BiGRU + ResNet50; double attention) w/ DropNet.
10	ORGANIZER	MMTen-ja	2020/08/26 22:52:22	3580	39.67	48.89	43.62	-	-	-	-	-	-	-	NMT	No	Baseline 3: multimodal NMT (Transformer + ResNet50 w/ visual attention)
11	TMU	MMTen-ja	2020/09/18 23:55:51	4020	39.32	48.59	43.30	-	-	-	-	-	-	-	NMT	No	Baseline: Multimodal NMT (BiGRU + ResNet50; decinit)
12	ORGANIZER	MMTen-ja	2020/08/26 22:50:13	3578	39.31	48.96	43.34	-	-	-	-	-	-	-	NMT	No	Baseline 5: multimodal NMT (Transformer + ResNet50 w/ visual attention and cross-lingual attention)
13	ORGANIZER	MMTen-ja	2020/08/26 22:47:43	3576	39.23	48.93	43.28	-	-	-	-	-	-	-	NMT	No	Baseline 1: text only NMT (Transformer)
14	ORGANIZER	MMTen-ja	2020/08/26 22:51:27	3579	39.17	48.75	43.15	-	-	-	-	-	-	-	NMT	No	Baseline 4: multimodal NMT (Transformer + ResNet50 w/ cross-lingual attention)
15	TMU	MMTen-ja	2020/09/18 19:33:49	3965	38.77	48.00	42.65	-	-	-	-	-	-	-	NMT	No	Multimodal NMT (BiGRU + ResNet50; double attention) w/ DropNet.
16	ORGANIZER	MMTen-ja	2020/08/26 22:49:09	3577	38.32	47.45	42.02	-	-	-	-	-	-	-	NMT	No	Baseline 2: multimodal NMT (Transformer + ResNet50)

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	TMEKU	MMTen-ja	2021/04/01 09:43:23	4730	0.886046	0.890507	0.886316	-	-	-	-	-	-	-	NMT	No	RNN-based MNMT model. Ensemble top 10 models.
2	sakura	MMTen-ja	2021/05/04 15:58:06	6256	0.883234	0.885923	0.882120	-	-	-	-	-	-	-	NMT	Yes	Text-only mBART25 finetuning
3	sakura	MMTen-ja	2021/05/04 18:29:15	6313	0.875110	0.879799	0.875825	-	-	-	-	-	-	-	NMT	No	The ensemble of two transformer-based MNMT models with universal visual representation
4	TMEKU	MMTen-ja	2021/04/23 22:24:24	5452	0.874392	0.880350	0.874700	-	-	-	-	-	-	-	NMT	No	RNN-based MNMT model. Single model.
5	TMU	MMTen-ja	2020/09/19 11:54:51	4046	0.871343	0.876282	0.871542	-	-	-	-	-	-	-	NMT	No	The Ensemble of 3 models. model: Multimodal NMT (BiGRU + ResNet50; double attention) w/ DropNet.
6	HW-TSC	MMTen-ja	2020/09/18 17:08:15	3914	0.870792	0.875362	0.870791	-	-	-	-	-	-	-	NMT	Yes	mmt_cmask_pretrained_transformer_big
7	TMU	MMTen-ja	2020/09/08 17:24:15	3651	0.869433	0.874742	0.869886	-	-	-	-	-	-	-	NMT	No	The Ensemble of top6 models.
8	TMU	MMTen-ja	2020/09/08 17:29:27	3653	0.868079	0.873126	0.868404	-	-	-	-	-	-	-	NMT	No	The Ensemble of 3 baselines. Baseline: Multimodal NMT (BiGRU + ResNet50; decinit)
9	TMU	MMTen-ja	2020/09/18 19:33:49	3965	0.865230	0.870127	0.865698	-	-	-	-	-	-	-	NMT	No	Multimodal NMT (BiGRU + ResNet50; double attention) w/ DropNet.
10	TMU	MMTen-ja	2020/09/18 23:55:51	4020	0.864242	0.868780	0.863866	-	-	-	-	-	-	-	NMT	No	Baseline: Multimodal NMT (BiGRU + ResNet50; decinit)
11	TMU	MMTen-ja	2020/09/15 15:21:42	3743	0.864209	0.870255	0.864792	-	-	-	-	-	-	-	NMT	No	Multimodal NMT (BiGRU + ResNet50; decinit) pretrained on noised F30kEnt-JP training data, fine-tuned on clean data.
12	ORGANIZER	MMTen-ja	2020/08/26 22:50:13	3578	0.863663	0.870539	0.865254	-	-	-	-	-	-	-	NMT	No	Baseline 5: multimodal NMT (Transformer + ResNet50 w/ visual attention and cross-lingual attention)
13	ORGANIZER	MMTen-ja	2020/08/26 22:51:27	3579	0.862288	0.868948	0.863400	-	-	-	-	-	-	-	NMT	No	Baseline 4: multimodal NMT (Transformer + ResNet50 w/ cross-lingual attention)
14	ORGANIZER	MMTen-ja	2020/08/26 22:47:43	3576	0.861623	0.869643	0.863487	-	-	-	-	-	-	-	NMT	No	Baseline 1: text only NMT (Transformer)
15	ORGANIZER	MMTen-ja	2020/08/26 22:52:22	3580	0.860180	0.867007	0.861681	-	-	-	-	-	-	-	NMT	No	Baseline 3: multimodal NMT (Transformer + ResNet50 w/ visual attention)
16	ORGANIZER	MMTen-ja	2020/08/26 22:49:09	3577	0.854899	0.862062	0.855312	-	-	-	-	-	-	-	NMT	No	Baseline 2: multimodal NMT (Transformer + ResNet50)

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	sakura	MMTen-ja	2021/05/04 15:58:06	6256	0.648369	0.648369	0.648369	-	-	-	-	-	-	-	NMT	Yes	Text-only mBART25 finetuning
2	sakura	MMTen-ja	2021/05/04 18:29:15	6313	0.644507	0.644507	0.644507	-	-	-	-	-	-	-	NMT	No	The ensemble of two transformer-based MNMT models with universal visual representation
3	TMEKU	MMTen-ja	2021/04/01 09:43:23	4730	0.644124	0.644124	0.644124	-	-	-	-	-	-	-	NMT	No	RNN-based MNMT model. Ensemble top 10 models.
4	TMEKU	MMTen-ja	2021/04/23 22:24:24	5452	0.644113	0.644113	0.644113	-	-	-	-	-	-	-	NMT	No	RNN-based MNMT model. Single model.
5	ORGANIZER	MMTen-ja	2020/08/26 22:52:22	3580	0.642177	0.642177	0.642177	-	-	-	-	-	-	-	NMT	No	Baseline 3: multimodal NMT (Transformer + ResNet50 w/ visual attention)
6	ORGANIZER	MMTen-ja	2020/08/26 22:49:09	3577	0.640381	0.640381	0.640381	-	-	-	-	-	-	-	NMT	No	Baseline 2: multimodal NMT (Transformer + ResNet50)
7	ORGANIZER	MMTen-ja	2020/08/26 22:51:27	3579	0.639181	0.639181	0.639181	-	-	-	-	-	-	-	NMT	No	Baseline 4: multimodal NMT (Transformer + ResNet50 w/ cross-lingual attention)
8	ORGANIZER	MMTen-ja	2020/08/26 22:50:13	3578	0.638311	0.638311	0.638311	-	-	-	-	-	-	-	NMT	No	Baseline 5: multimodal NMT (Transformer + ResNet50 w/ visual attention and cross-lingual attention)
9	ORGANIZER	MMTen-ja	2020/08/26 22:47:43	3576	0.637608	0.637608	0.637608	-	-	-	-	-	-	-	NMT	No	Baseline 1: text only NMT (Transformer)
10	TMU	MMTen-ja	2020/09/08 17:24:15	3651	0.000000	0.000000	0.000000	-	-	-	-	-	-	-	NMT	No	The Ensemble of top6 models.
11	TMU	MMTen-ja	2020/09/08 17:29:27	3653	0.000000	0.000000	0.000000	-	-	-	-	-	-	-	NMT	No	The Ensemble of 3 baselines. Baseline: Multimodal NMT (BiGRU + ResNet50; decinit)
12	TMU	MMTen-ja	2020/09/15 15:21:42	3743	0.000000	0.000000	0.000000	-	-	-	-	-	-	-	NMT	No	Multimodal NMT (BiGRU + ResNet50; decinit) pretrained on noised F30kEnt-JP training data, fine-tuned on clean data.
13	HW-TSC	MMTen-ja	2020/09/18 17:08:15	3914	0.000000	0.000000	0.000000	-	-	-	-	-	-	-	NMT	Yes	mmt_cmask_pretrained_transformer_big
14	TMU	MMTen-ja	2020/09/18 19:33:49	3965	0.000000	0.000000	0.000000	-	-	-	-	-	-	-	NMT	No	Multimodal NMT (BiGRU + ResNet50; double attention) w/ DropNet.
15	TMU	MMTen-ja	2020/09/18 23:55:51	4020	0.000000	0.000000	0.000000	-	-	-	-	-	-	-	NMT	No	Baseline: Multimodal NMT (BiGRU + ResNet50; decinit)
16	TMU	MMTen-ja	2020/09/19 11:54:51	4046	0.000000	0.000000	0.000000	-	-	-	-	-	-	-	NMT	No	The Ensemble of 3 models. model: Multimodal NMT (BiGRU + ResNet50; double attention) w/ DropNet.

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