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	cvit	HINDENen-hi	2018/09/18 22:37:37	2500	-	-	-	-	-	-	21.57	-	0.00	0.00	NMT	Yes	Averaging Models from epochs 61-68. Base Transformer. Uses External Data.
2	XMUNLP	HINDENen-hi	2017/07/28 23:38:29	1576	-	-	-	-	-	-	21.39	0.00	0.00	0.00	NMT	No	ensemble of 4 nmt models + monolingual data
3	cvit	HINDENen-hi	2018/09/18 21:58:21	2496	-	-	-	-	-	-	21.35	-	0.00	0.00	NMT	Yes	Transformer Base. Uses External Data. Averaging of Checkpoints Enabled.
4	cvit	HINDENen-hi	2018/09/18 15:21:13	2489	-	-	-	-	-	-	21.10	-	0.00	0.00	NMT	Yes	Transformer Base. Uses External Data
5	cvit	HINDENen-hi	2019/05/27 16:03:36	2680	-	-	-	-	-	-	20.46	-	0.00	0.00	NMT	Yes	massive-multi + bt
6	CUNI	HINDENen-hi	2018/09/15 01:12:40	2361	-	-	-	-	-	-	20.28	-	0.00	0.00	NMT	No	Transformer big, only backtranslation EN-HI, no original EN-HI, beam=8; alpha=0.8; averaging of last 8 models after 1300k steps
7	cvit	HINDENen-hi	2019/03/15 01:21:27	2642	-	-	-	-	-	-	20.17	-	0.00	0.00	NMT	Yes	massive-multi
8	CUNI	HINDENen-hi	2018/09/15 01:22:03	2365	-	-	-	-	-	-	20.07	-	0.00	0.00	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, only backtranslation EN-HI, no original EN-HI, beam=8; alpha=0.8; averaging of last 8 models after 700k steps
9	XMUNLP	HINDENen-hi	2017/07/27 22:04:54	1508	-	-	-	-	-	-	19.79	0.00	0.00	0.00	NMT	No	single nmt model + monolingual data
10	CUNI	HINDENen-hi	2018/09/13 22:18:14	2320	-	-	-	-	-	-	19.78	-	0.00	0.00	NMT	No	Big Transformer model with backtranslation, with transfer learning from English to Czech.
11	cvit	HINDENen-hi	2018/09/09 21:12:29	2254	-	-	-	-	-	-	19.69	-	0.00	0.00	NMT	Yes	ConvS2S. Uses external data.
12	cvit	HINDENen-hi	2018/09/07 12:29:04	2235	-	-	-	-	-	-	18.77	-	0.00	0.00	NMT	Yes	ConvS2S Model. External Data is used.
13	ORGANIZER	HINDENen-hi	2016/07/26 10:07:48	1032	-	-	-	-	-	-	18.72	0.00	0.00	0.00	Other	Yes	Online A (2016)
14	cvit	HINDENen-hi	2019/03/15 01:31:41	2644	-	-	-	-	-	-	18.31	-	0.00	0.00	NMT	Yes	massive-multi + ft
15	CUNI	HINDENen-hi	2018/09/15 01:14:34	2362	-	-	-	-	-	-	17.63	-	0.00	0.00	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, followed by only backtranslation EN-HI for 300k steps, followed by original EN-HI for 500k steps, beam=8; alpha=0.8; averaging of last 8 models.
16	ORGANIZER	HINDENen-hi	2016/07/26 13:24:22	1047	-	-	-	-	-	-	16.97	0.00	0.00	0.00	Other	Yes	Online B (2016)
17	cvit	HINDENen-hi	2018/09/09 01:20:09	2251	-	-	-	-	-	-	16.77	-	0.00	0.00	NMT	No	ConvS2S Model. IIT-Bombay data filtered with langdetect. + Backtranslated Monolingual Data ppl in [0.05, 0.14]
18	CUNI	HINDENen-hi	2018/09/15 01:19:04	2363	-	-	-	-	-	-	16.49	-	0.00	0.00	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, only original EN-HI, beam=8; alpha=0.8; averaging of last 8 models after 230k steps.
19	CUNI	HINDENen-hi	2018/09/15 01:20:33	2364	-	-	-	-	-	-	14.20	-	0.00	0.00	NMT	No	Baseline, transformer big only EN-HI, beam=8, alpha=0.8, averaging 8 steps after 330k steps
20	ORGANIZER	HINDENen-hi	2018/11/13 14:54:58	2566	-	-	-	-	-	-	13.76	-	0.00	0.00	NMT	No	NMT with Attention
21	IITP-MT	HINDENen-hi	2016/08/18 23:13:25	1185	-	-	-	-	-	-	13.71	0.00	0.00	0.00	SMT	Yes	IITP-MT System1
22	XMUNLP	HINDENen-hi	2017/07/20 23:07:38	1422	-	-	-	-	-	-	13.69	0.00	0.00	0.00	NMT	No	single nmt model
23	IITP-MT	HINDENen-hi	2016/08/29 18:51:44	1290	-	-	-	-	-	-	13.57	0.00	0.00	0.00	SMT	No	IITP-MT System2
24	IITB-MTG	HINDENen-hi	2017/08/01 15:09:01	1725	-	-	-	-	-	-	12.23	0.00	0.00	0.00	NMT	No	NMT with ensemble (last 3 + best validation)
25	EHR	HINDENen-hi	2016/08/17 14:30:08	1166	-	-	-	-	-	-	11.75	0.00	0.00	0.00	SMT	No	PBSMT with preordering (DL=6)
26	ORGANIZER	HINDENen-hi	2016/08/20 17:41:36	1252	-	-	-	-	-	-	10.79	0.00	0.00	0.00	SMT	No	Phrase-based SMT
27	IITB-MTG	HINDENen-hi	2017/09/05 23:04:58	1763	-	-	-	-	-	-	0.34	0.00	0.00	0.00	NMT	No

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	cvit	HINDENen-hi	2018/09/18 22:37:37	2500	-	-	-	-	-	-	0.773923	-	0.000000	0.000000	NMT	Yes	Averaging Models from epochs 61-68. Base Transformer. Uses External Data.
2	cvit	HINDENen-hi	2018/09/18 21:58:21	2496	-	-	-	-	-	-	0.773078	-	0.000000	0.000000	NMT	Yes	Transformer Base. Uses External Data. Averaging of Checkpoints Enabled.
3	cvit	HINDENen-hi	2018/09/18 15:21:13	2489	-	-	-	-	-	-	0.771549	-	0.000000	0.000000	NMT	Yes	Transformer Base. Uses External Data
4	cvit	HINDENen-hi	2019/05/27 16:03:36	2680	-	-	-	-	-	-	0.765422	-	0.000000	0.000000	NMT	Yes	massive-multi + bt
5	CUNI	HINDENen-hi	2018/09/15 01:22:03	2365	-	-	-	-	-	-	0.761582	-	0.000000	0.000000	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, only backtranslation EN-HI, no original EN-HI, beam=8; alpha=0.8; averaging of last 8 models after 700k steps
6	CUNI	HINDENen-hi	2018/09/15 01:12:40	2361	-	-	-	-	-	-	0.761292	-	0.000000	0.000000	NMT	No	Transformer big, only backtranslation EN-HI, no original EN-HI, beam=8; alpha=0.8; averaging of last 8 models after 1300k steps
7	cvit	HINDENen-hi	2019/03/15 01:21:27	2642	-	-	-	-	-	-	0.761061	-	0.000000	0.000000	NMT	Yes	massive-multi
8	cvit	HINDENen-hi	2018/09/09 21:12:29	2254	-	-	-	-	-	-	0.758365	-	0.000000	0.000000	NMT	Yes	ConvS2S. Uses external data.
9	CUNI	HINDENen-hi	2018/09/15 01:19:04	2363	-	-	-	-	-	-	0.754966	-	0.000000	0.000000	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, only original EN-HI, beam=8; alpha=0.8; averaging of last 8 models after 230k steps.
10	CUNI	HINDENen-hi	2018/09/13 22:18:14	2320	-	-	-	-	-	-	0.754244	-	0.000000	0.000000	NMT	No	Big Transformer model with backtranslation, with transfer learning from English to Czech.
11	CUNI	HINDENen-hi	2018/09/15 01:14:34	2362	-	-	-	-	-	-	0.753895	-	0.000000	0.000000	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, followed by only backtranslation EN-HI for 300k steps, followed by original EN-HI for 500k steps, beam=8; alpha=0.8; averaging of last 8 models.
12	XMUNLP	HINDENen-hi	2017/07/28 23:38:29	1576	-	-	-	-	-	-	0.749660	0.000000	0.000000	0.000000	NMT	No	ensemble of 4 nmt models + monolingual data
13	cvit	HINDENen-hi	2018/09/07 12:29:04	2235	-	-	-	-	-	-	0.748008	-	0.000000	0.000000	NMT	Yes	ConvS2S Model. External Data is used.
14	XMUNLP	HINDENen-hi	2017/07/27 22:04:54	1508	-	-	-	-	-	-	0.743129	0.000000	0.000000	0.000000	NMT	No	single nmt model + monolingual data
15	CUNI	HINDENen-hi	2018/09/15 01:20:33	2364	-	-	-	-	-	-	0.733738	-	0.000000	0.000000	NMT	No	Baseline, transformer big only EN-HI, beam=8, alpha=0.8, averaging 8 steps after 330k steps
16	cvit	HINDENen-hi	2019/03/15 01:31:41	2644	-	-	-	-	-	-	0.718374	-	0.000000	0.000000	NMT	Yes	massive-multi + ft
17	ORGANIZER	HINDENen-hi	2016/07/26 10:07:48	1032	-	-	-	-	-	-	0.716788	0.000000	0.000000	0.000000	Other	Yes	Online A (2016)
18	cvit	HINDENen-hi	2018/09/09 01:20:09	2251	-	-	-	-	-	-	0.714197	-	0.000000	0.000000	NMT	No	ConvS2S Model. IIT-Bombay data filtered with langdetect. + Backtranslated Monolingual Data ppl in [0.05, 0.14]
19	XMUNLP	HINDENen-hi	2017/07/20 23:07:38	1422	-	-	-	-	-	-	0.712876	0.000000	0.000000	0.000000	NMT	No	single nmt model
20	ORGANIZER	HINDENen-hi	2018/11/13 14:54:58	2566	-	-	-	-	-	-	0.710210	-	0.000000	0.000000	NMT	No	NMT with Attention
21	ORGANIZER	HINDENen-hi	2016/07/26 13:24:22	1047	-	-	-	-	-	-	0.691298	0.000000	0.000000	0.000000	Other	Yes	Online B (2016)
22	IITP-MT	HINDENen-hi	2016/08/18 23:13:25	1185	-	-	-	-	-	-	0.688913	0.000000	0.000000	0.000000	SMT	Yes	IITP-MT System1
23	IITB-MTG	HINDENen-hi	2017/08/01 15:09:01	1725	-	-	-	-	-	-	0.688606	0.000000	0.000000	0.000000	NMT	No	NMT with ensemble (last 3 + best validation)
24	IITP-MT	HINDENen-hi	2016/08/29 18:51:44	1290	-	-	-	-	-	-	0.683022	0.000000	0.000000	0.000000	SMT	No	IITP-MT System2
25	EHR	HINDENen-hi	2016/08/17 14:30:08	1166	-	-	-	-	-	-	0.671866	0.000000	0.000000	0.000000	SMT	No	PBSMT with preordering (DL=6)
26	ORGANIZER	HINDENen-hi	2016/08/20 17:41:36	1252	-	-	-	-	-	-	0.651166	0.000000	0.000000	0.000000	SMT	No	Phrase-based SMT
27	IITB-MTG	HINDENen-hi	2017/09/05 23:04:58	1763	-	-	-	-	-	-	0.301241	0.000000	0.000000	0.000000	NMT	No

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	juman	kytea	mecab	moses- tokenizer	stanford- segmenter- ctb	stanford- segmenter- pku	indic- tokenizer	unuse	myseg	kmseg	Method	Other Resources	System Description
1	cvit	HINDENen-hi	2018/09/18 15:21:13	2489	-	-	-	-	-	-	0.712200	-	0.000000	0.000000	NMT	Yes	Transformer Base. Uses External Data
2	cvit	HINDENen-hi	2018/09/18 22:37:37	2500	-	-	-	-	-	-	0.712110	-	0.000000	0.000000	NMT	Yes	Averaging Models from epochs 61-68. Base Transformer. Uses External Data.
3	cvit	HINDENen-hi	2018/09/18 21:58:21	2496	-	-	-	-	-	-	0.712010	-	0.000000	0.000000	NMT	Yes	Transformer Base. Uses External Data. Averaging of Checkpoints Enabled.
4	CUNI	HINDENen-hi	2018/09/15 01:12:40	2361	-	-	-	-	-	-	0.704220	-	0.000000	0.000000	NMT	No	Transformer big, only backtranslation EN-HI, no original EN-HI, beam=8; alpha=0.8; averaging of last 8 models after 1300k steps
5	cvit	HINDENen-hi	2019/05/27 16:03:36	2680	-	-	-	-	-	-	0.702380	-	0.000000	0.000000	NMT	Yes	massive-multi + bt
6	cvit	HINDENen-hi	2019/03/15 01:21:27	2642	-	-	-	-	-	-	0.701670	-	0.000000	0.000000	NMT	Yes	massive-multi
7	CUNI	HINDENen-hi	2018/09/15 01:22:03	2365	-	-	-	-	-	-	0.701300	-	0.000000	0.000000	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, only backtranslation EN-HI, no original EN-HI, beam=8; alpha=0.8; averaging of last 8 models after 700k steps
8	CUNI	HINDENen-hi	2018/09/13 22:18:14	2320	-	-	-	-	-	-	0.700240	-	0.000000	0.000000	NMT	No	Big Transformer model with backtranslation, with transfer learning from English to Czech.
9	cvit	HINDENen-hi	2018/09/09 21:12:29	2254	-	-	-	-	-	-	0.699810	-	0.000000	0.000000	NMT	Yes	ConvS2S. Uses external data.
10	cvit	HINDENen-hi	2018/09/07 12:29:04	2235	-	-	-	-	-	-	0.697630	-	0.000000	0.000000	NMT	Yes	ConvS2S Model. External Data is used.
11	CUNI	HINDENen-hi	2018/09/15 01:14:34	2362	-	-	-	-	-	-	0.693830	-	0.000000	0.000000	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, followed by only backtranslation EN-HI for 300k steps, followed by original EN-HI for 500k steps, beam=8; alpha=0.8; averaging of last 8 models.
12	CUNI	HINDENen-hi	2018/09/15 01:19:04	2363	-	-	-	-	-	-	0.690150	-	0.000000	0.000000	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, only original EN-HI, beam=8; alpha=0.8; averaging of last 8 models after 230k steps.
13	XMUNLP	HINDENen-hi	2017/07/28 23:38:29	1576	-	-	-	-	-	-	0.688770	0.000000	0.000000	0.000000	NMT	No	ensemble of 4 nmt models + monolingual data
14	XMUNLP	HINDENen-hi	2017/07/27 22:04:54	1508	-	-	-	-	-	-	0.682500	0.000000	0.000000	0.000000	NMT	No	single nmt model + monolingual data
15	CUNI	HINDENen-hi	2018/09/15 01:20:33	2364	-	-	-	-	-	-	0.681460	-	0.000000	0.000000	NMT	No	Baseline, transformer big only EN-HI, beam=8, alpha=0.8, averaging 8 steps after 330k steps
16	cvit	HINDENen-hi	2019/03/15 01:31:41	2644	-	-	-	-	-	-	0.680620	-	0.000000	0.000000	NMT	Yes	massive-multi + ft
17	ORGANIZER	HINDENen-hi	2016/07/26 10:07:48	1032	-	-	-	-	-	-	0.670660	0.000000	0.000000	0.000000	Other	Yes	Online A (2016)
18	ORGANIZER	HINDENen-hi	2016/07/26 13:24:22	1047	-	-	-	-	-	-	0.668450	0.000000	0.000000	0.000000	Other	Yes	Online B (2016)
19	cvit	HINDENen-hi	2018/09/09 01:20:09	2251	-	-	-	-	-	-	0.664330	-	0.000000	0.000000	NMT	No	ConvS2S Model. IIT-Bombay data filtered with langdetect. + Backtranslated Monolingual Data ppl in [0.05, 0.14]
20	IITP-MT	HINDENen-hi	2016/08/29 18:51:44	1290	-	-	-	-	-	-	0.663210	0.000000	0.000000	0.000000	SMT	No	IITP-MT System2
21	ORGANIZER	HINDENen-hi	2016/08/20 17:41:36	1252	-	-	-	-	-	-	0.660860	0.000000	0.000000	0.000000	SMT	No	Phrase-based SMT
22	IITP-MT	HINDENen-hi	2016/08/18 23:13:25	1185	-	-	-	-	-	-	0.657330	0.000000	0.000000	0.000000	SMT	Yes	IITP-MT System1
23	EHR	HINDENen-hi	2016/08/17 14:30:08	1166	-	-	-	-	-	-	0.650750	0.000000	0.000000	0.000000	SMT	No	PBSMT with preordering (DL=6)
24	XMUNLP	HINDENen-hi	2017/07/20 23:07:38	1422	-	-	-	-	-	-	0.647740	0.000000	0.000000	0.000000	NMT	No	single nmt model
25	ORGANIZER	HINDENen-hi	2018/11/13 14:54:58	2566	-	-	-	-	-	-	0.644860	-	0.000000	0.000000	NMT	No	NMT with Attention
26	IITB-MTG	HINDENen-hi	2017/08/01 15:09:01	1725	-	-	-	-	-	-	0.624780	0.000000	0.000000	0.000000	NMT	No	NMT with ensemble (last 3 + best validation)
27	IITB-MTG	HINDENen-hi	2017/09/05 23:04:58	1763	-	-	-	-	-	-	0.463350	0.000000	0.000000	0.000000	NMT	No

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.
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 (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	CUNI	HINDENen-hi	2018/09/15 01:14:34	2362	77.000	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, followed by only backtranslation EN-HI for 300k steps, followed by original EN-HI for 500k steps, beam=8; alpha=0.8; averaging of last 8 models.
2	cvit	HINDENen-hi	2018/09/09 21:12:29	2254	69.500	NMT	Yes	ConvS2S. Uses external data.
3	CUNI	HINDENen-hi	2018/09/15 01:22:03	2365	60.000	NMT	No	Transformer big, transfer learning from EN-CS 1M steps, only backtranslation EN-HI, no original EN-HI, beam=8; alpha=0.8; averaging of last 8 models after 700k steps
4	cvit	HINDENen-hi	2018/09/09 01:20:09	2251	50.500	NMT	No	ConvS2S Model. IIT-Bombay data filtered with langdetect. + Backtranslated Monolingual Data ppl in [0.05, 0.14]
5	cvit	HINDENen-hi	2018/09/07 12:29:04	2235	Underway	NMT	Yes	ConvS2S Model. External Data is used.

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 (WAT2019)

HUMAN (WAT2018)

HUMAN (WAT2017)

HUMAN (WAT2016)

HUMAN (WAT2015)

HUMAN (WAT2014)

EVALUATION RESULTS USAGE POLICY

WAT

The Workshop on Asian Translation

Evaluation Results