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	NHK-NES	JIJIja-en	2019/07/24 14:03:10	2883	-	-	-	26.38	-	-	-	-	-	-	NMT	Yes	ensemble-10, beamwidth-30 with other resources (including BT)
2	sarah	JIJIja-en	2019/07/22 11:59:11	2793	-	-	-	21.84	-	-	-	-	-	-	NMT	No	Transformer, ensemble of 4 models
3	sarah	JIJIja-en	2019/07/22 16:41:43	2813	-	-	-	21.34	-	-	-	-	-	-	NMT	No	Transformer, single model
4	NTT	JIJIja-en	2017/08/01 02:55:31	1677	-	-	-	20.90	-	-	-	0.00	0.00	0.00	NMT	No	Ensemble 8 Models: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 20 w/ length-based reranking
5	NTT	JIJIja-en	2017/07/30 10:16:39	1599	-	-	-	19.44	-	-	-	0.00	0.00	0.00	NMT	No	Single Model: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 20 w/ length-based reranking
6	NICT-2	JIJIja-en	2017/07/26 13:49:39	1474	-	-	-	18.19	-	-	-	0.00	0.00	0.00	NMT	No	NMT 16 Ensembles * Bi-directional Reranking
7	XMUNLP	JIJIja-en	2017/07/24 20:42:39	1442	-	-	-	17.95	-	-	-	0.00	0.00	0.00	NMT	No	ensemble of 4 nmt models
8	NICT-2	JIJIja-en	2017/07/26 13:46:11	1473	-	-	-	16.52	-	-	-	0.00	0.00	0.00	NMT	No	NMT Single Model: BPE35k, Bi-LSTM(500*2) Encoder, LSTM(1000) Left-to-Right Decoder
9	ORGANIZER	JIJIja-en	2018/08/14 11:49:18	1905	-	-	-	16.48	-	-	-	-	0.00	0.00	NMT	No	NMT with Attention
10	XMUNLP	JIJIja-en	2017/07/24 09:02:37	1428	-	-	-	15.77	-	-	-	0.00	0.00	0.00	NMT	No	single nmt model
11	NTT	JIJIja-en	2017/07/28 19:24:02	1564	-	-	-	15.77	-	-	-	0.00	0.00	0.00	NMT	Yes	Single Model: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 5 w/ length-based reranking / initialized model = ASPEC + JPO
12	ORGANIZER	JIJIja-en	2017/07/19 13:12:43	1396	-	-	-	15.67	-	-	-	0.00	0.00	0.00	SMT	No	Hierarchical Phrase-based SMT
13	ORGANIZER	JIJIja-en	2017/07/19 12:51:03	1394	-	-	-	15.11	-	-	-	0.00	0.00	0.00	SMT	No	Phrase-based SMT
14	ORGANIZER	JIJIja-en	2017/07/19 13:39:22	1398	-	-	-	14.54	-	-	-	0.00	0.00	0.00	SMT	No	String-to-Tree SMT
15	NHK-NES	JIJIja-en	2019/07/24 14:07:56	2884	-	-	-	14.23	-	-	-	-	-	-	NMT	Yes	ensemble-10, beamwidth-30 using another domain-tag (including BT)
16	CUNI	JIJIja-en	2017/07/31 22:38:53	1668	-	-	-	10.67	-	-	-	0.00	0.00	0.00	SMT	No	Bahdanau (2014) seq2seq with conditional GRU on byte-pair encoding
17	ORGANIZER	JIJIja-en	2017/07/28 15:24:58	1523	-	-	-	8.19	-	-	-	0.00	0.00	0.00	NMT	No	ONLINE-A
18	ORGANIZER	JIJIja-en	2017/07/28 15:25:47	1524	-	-	-	5.17	-	-	-	0.00	0.00	0.00	NMT	No	ONLINE-B
19	ORGANIZER	JIJIja-en	2017/07/28 15:27:39	1526	-	-	-	4.67	-	-	-	0.00	0.00	0.00	RBMT	No	RBMT-B
20	ORGANIZER	JIJIja-en	2017/07/28 15:26:56	1525	-	-	-	4.36	-	-	-	0.00	0.00	0.00	RBMT	No	RBMT-A

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	NHK-NES	JIJIja-en	2019/07/24 14:03:10	2883	-	-	-	0.703808	-	-	-	-	-	-	NMT	Yes	ensemble-10, beamwidth-30 with other resources (including BT)
2	sarah	JIJIja-en	2019/07/22 16:41:43	2813	-	-	-	0.676723	-	-	-	-	-	-	NMT	No	Transformer, single model
3	sarah	JIJIja-en	2019/07/22 11:59:11	2793	-	-	-	0.675386	-	-	-	-	-	-	NMT	No	Transformer, ensemble of 4 models
4	NTT	JIJIja-en	2017/08/01 02:55:31	1677	-	-	-	0.648931	-	-	-	0.000000	0.000000	0.000000	NMT	No	Ensemble 8 Models: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 20 w/ length-based reranking
5	NICT-2	JIJIja-en	2017/07/26 13:46:11	1473	-	-	-	0.642379	-	-	-	0.000000	0.000000	0.000000	NMT	No	NMT Single Model: BPE35k, Bi-LSTM(500*2) Encoder, LSTM(1000) Left-to-Right Decoder
6	ORGANIZER	JIJIja-en	2018/08/14 11:49:18	1905	-	-	-	0.640558	-	-	-	-	0.000000	0.000000	NMT	No	NMT with Attention
7	NTT	JIJIja-en	2017/07/30 10:16:39	1599	-	-	-	0.638841	-	-	-	0.000000	0.000000	0.000000	NMT	No	Single Model: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 20 w/ length-based reranking
8	XMUNLP	JIJIja-en	2017/07/24 20:42:39	1442	-	-	-	0.637059	-	-	-	0.000000	0.000000	0.000000	NMT	No	ensemble of 4 nmt models
9	NICT-2	JIJIja-en	2017/07/26 13:49:39	1474	-	-	-	0.632638	-	-	-	0.000000	0.000000	0.000000	NMT	No	NMT 16 Ensembles * Bi-directional Reranking
10	NTT	JIJIja-en	2017/07/28 19:24:02	1564	-	-	-	0.627182	-	-	-	0.000000	0.000000	0.000000	NMT	Yes	Single Model: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 5 w/ length-based reranking / initialized model = ASPEC + JPO
11	XMUNLP	JIJIja-en	2017/07/24 09:02:37	1428	-	-	-	0.617123	-	-	-	0.000000	0.000000	0.000000	NMT	No	single nmt model
12	NHK-NES	JIJIja-en	2019/07/24 14:07:56	2884	-	-	-	0.612351	-	-	-	-	-	-	NMT	Yes	ensemble-10, beamwidth-30 using another domain-tag (including BT)
13	CUNI	JIJIja-en	2017/07/31 22:38:53	1668	-	-	-	0.564797	-	-	-	0.000000	0.000000	0.000000	SMT	No	Bahdanau (2014) seq2seq with conditional GRU on byte-pair encoding
14	ORGANIZER	JIJIja-en	2017/07/19 13:12:43	1396	-	-	-	0.558225	-	-	-	0.000000	0.000000	0.000000	SMT	No	Hierarchical Phrase-based SMT
15	ORGANIZER	JIJIja-en	2017/07/19 13:39:22	1398	-	-	-	0.556728	-	-	-	0.000000	0.000000	0.000000	SMT	No	String-to-Tree SMT
16	ORGANIZER	JIJIja-en	2017/07/19 12:51:03	1394	-	-	-	0.554550	-	-	-	0.000000	0.000000	0.000000	SMT	No	Phrase-based SMT
17	ORGANIZER	JIJIja-en	2017/07/28 15:24:58	1523	-	-	-	0.529844	-	-	-	0.000000	0.000000	0.000000	NMT	No	ONLINE-A
18	ORGANIZER	JIJIja-en	2017/07/28 15:27:39	1526	-	-	-	0.475760	-	-	-	0.000000	0.000000	0.000000	RBMT	No	RBMT-B
19	ORGANIZER	JIJIja-en	2017/07/28 15:26:56	1525	-	-	-	0.472312	-	-	-	0.000000	0.000000	0.000000	RBMT	No	RBMT-A
20	ORGANIZER	JIJIja-en	2017/07/28 15:25:47	1524	-	-	-	0.461632	-	-	-	0.000000	0.000000	0.000000	NMT	No	ONLINE-B

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	NHK-NES	JIJIja-en	2019/07/24 14:03:10	2883	-	-	-	0.554110	-	-	-	-	-	-	NMT	Yes	ensemble-10, beamwidth-30 with other resources (including BT)
2	sarah	JIJIja-en	2019/07/22 11:59:11	2793	-	-	-	0.526530	-	-	-	-	-	-	NMT	No	Transformer, ensemble of 4 models
3	NHK-NES	JIJIja-en	2019/07/24 14:07:56	2884	-	-	-	0.526010	-	-	-	-	-	-	NMT	Yes	ensemble-10, beamwidth-30 using another domain-tag (including BT)
4	sarah	JIJIja-en	2019/07/22 16:41:43	2813	-	-	-	0.524530	-	-	-	-	-	-	NMT	No	Transformer, single model
5	ORGANIZER	JIJIja-en	2017/07/19 13:39:22	1398	-	-	-	0.477170	-	-	-	0.000000	0.000000	0.000000	SMT	No	String-to-Tree SMT
6	NTT	JIJIja-en	2017/07/30 10:16:39	1599	-	-	-	0.476200	-	-	-	0.000000	0.000000	0.000000	NMT	No	Single Model: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 20 w/ length-based reranking
7	ORGANIZER	JIJIja-en	2017/07/19 12:51:03	1394	-	-	-	0.475740	-	-	-	0.000000	0.000000	0.000000	SMT	No	Phrase-based SMT
8	NTT	JIJIja-en	2017/08/01 02:55:31	1677	-	-	-	0.474360	-	-	-	0.000000	0.000000	0.000000	NMT	No	Ensemble 8 Models: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 20 w/ length-based reranking
9	NTT	JIJIja-en	2017/07/28 19:24:02	1564	-	-	-	0.473970	-	-	-	0.000000	0.000000	0.000000	NMT	Yes	Single Model: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 5 w/ length-based reranking / initialized model = ASPEC + JPO
10	ORGANIZER	JIJIja-en	2017/07/19 13:12:43	1396	-	-	-	0.470610	-	-	-	0.000000	0.000000	0.000000	SMT	No	Hierarchical Phrase-based SMT
11	XMUNLP	JIJIja-en	2017/07/24 20:42:39	1442	-	-	-	0.465710	-	-	-	0.000000	0.000000	0.000000	NMT	No	ensemble of 4 nmt models
12	XMUNLP	JIJIja-en	2017/07/24 09:02:37	1428	-	-	-	0.463550	-	-	-	0.000000	0.000000	0.000000	NMT	No	single nmt model
13	ORGANIZER	JIJIja-en	2018/08/14 11:49:18	1905	-	-	-	0.459080	-	-	-	-	0.000000	0.000000	NMT	No	NMT with Attention
14	NICT-2	JIJIja-en	2017/07/26 13:46:11	1473	-	-	-	0.459000	-	-	-	0.000000	0.000000	0.000000	NMT	No	NMT Single Model: BPE35k, Bi-LSTM(500*2) Encoder, LSTM(1000) Left-to-Right Decoder
15	NICT-2	JIJIja-en	2017/07/26 13:49:39	1474	-	-	-	0.453420	-	-	-	0.000000	0.000000	0.000000	NMT	No	NMT 16 Ensembles * Bi-directional Reranking
16	ORGANIZER	JIJIja-en	2017/07/28 15:24:58	1523	-	-	-	0.450850	-	-	-	0.000000	0.000000	0.000000	NMT	No	ONLINE-A
17	CUNI	JIJIja-en	2017/07/31 22:38:53	1668	-	-	-	0.432700	-	-	-	0.000000	0.000000	0.000000	SMT	No	Bahdanau (2014) seq2seq with conditional GRU on byte-pair encoding
18	ORGANIZER	JIJIja-en	2017/07/28 15:26:56	1525	-	-	-	0.391050	-	-	-	0.000000	0.000000	0.000000	RBMT	No	RBMT-A
19	ORGANIZER	JIJIja-en	2017/07/28 15:27:39	1526	-	-	-	0.385600	-	-	-	0.000000	0.000000	0.000000	RBMT	No	RBMT-B
20	ORGANIZER	JIJIja-en	2017/07/28 15:25:47	1524	-	-	-	0.367760	-	-	-	0.000000	0.000000	0.000000	NMT	No	ONLINE-B

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)

#	Team	Task	Date/Time	DataID	HUMAN	Method	Other Resources	System Description
1	ORGANIZER	JIJIja-en	2017/07/28 15:24:58	1523	70.000	NMT	No	ONLINE-A
2	ORGANIZER	JIJIja-en	2017/07/28 15:27:39	1526	51.750	RBMT	No	RBMT-B
3	NTT	JIJIja-en	2017/07/30 10:16:39	1599	32.000	NMT	No	Single Model: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 20 w/ length-based reranking
4	NTT	JIJIja-en	2017/08/01 02:55:31	1677	26.750	NMT	No	Ensemble 8 Models: joint BPE 16k, BiLSTM Encoder 51222, LtoR LSTM Decoder 512*2, Beam Search 20 w/ length-based reranking
5	XMUNLP	JIJIja-en	2017/07/24 20:42:39	1442	20.750	NMT	No	ensemble of 4 nmt models
6	ORGANIZER	JIJIja-en	2017/07/19 13:12:43	1396	10.250	SMT	No	Hierarchical Phrase-based SMT
7	NICT-2	JIJIja-en	2017/07/26 13:49:39	1474	7.250	NMT	No	NMT 16 Ensembles * Bi-directional Reranking
8	NICT-2	JIJIja-en	2017/07/26 13:46:11	1473	0.250	NMT	No	NMT Single Model: BPE35k, Bi-LSTM(500*2) Encoder, LSTM(1000) Left-to-Right Decoder
9	CUNI	JIJIja-en	2017/07/31 22:38:53	1668	-24.000	SMT	No	Bahdanau (2014) seq2seq with conditional GRU on byte-pair encoding

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