NICT_LOGO.JPG KYOTO-U_LOGO.JPG

WAT

The Workshop on Asian Translation
Evaluation Results

[EVALUATION RESULTS TOP] | [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]

BLEU


# Team Task Date/Time DataID BLEU
Method
 Other
Resources
 System
Description
juman kytea mecab moses-
tokenizer		 stanford-
segmenter-
ctb		 stanford-
segmenter-
pku		 indic-
tokenizer		 unuse myseg kmseg
100-7MMCHMM24en-hi2024/08/05 15:02:387190------53.40---NMTYeschallenge set
2BITS-PMMCHMM24en-hi2023/07/08 13:44:397124------52.10---NMTYesNLLB model finetuned on captions + object tags of original & synthetic images using DETR model
3VoltaMMCHMM24en-hi2021/05/25 13:56:036430------51.60---NMTYesFinetuned mBART (Used IITB for data augmentation) and added object tags to the input using Mask RCNN
4v036MMCHMM24en-hi2024/08/15 18:19:507406------43.20---NMTNo
5ODIAGENMMCHMM24en-hi2023/07/06 03:54:047106------42.80---NMTNoImage features extracted as Object tags appended with text and MBART fine-tuning
6v036MMCHMM24en-hi2024/08/11 12:43:407319------42.50---NMTNoNMT based system using both image descriptors and text description. A multistage LLM pipeline used for extracting image data descriptions and translation. Fine tuning done in few cases Models Used:
7v036MMCHMM24en-hi2024/08/11 23:48:237353------40.30---NMTNo
8v036MMCHMM24en-hi2024/08/14 20:25:457397------39.70---NMTNo
9CNLP-NITS-PPMMCHMM24en-hi2022/07/11 12:39:256741------39.30---NMTNoTransliteration-based phrase pairs augmentation and visual features in training using BRNN encoder and doubly-attentive-rnn decoder.
10CNLP-NITS-PPMMCHMM24en-hi2021/04/27 23:56:015730------39.28---NMTYesPretrained monolingual data (IITB) using Glove and fine-tuned with parallel data (WAT21 train data+ Extracted Phrase pairs from WAT21 train data +IITB train data) and visual features in training using
11SILO_NLPMMCHMM24en-hi2022/07/18 17:33:086959------39.10---NMTYesObject Tags (Image) + Flickr8 dataset as additional resource + Finetune mBART
12239233MMCHMM24en-hi2024/08/13 07:05:337378------37.90---NMTYesOne-shot prompt for synthetic QA description from captions; translate QA using IndicTrans2; generate caption from QA as context
13iitpMMCHMM24en-hi2021/05/01 23:25:565942------37.50---NMTNoRemoved special chars at start and end of sentence 1. pre-trained with HindEnCorp, trained with Vis Gen 2. trained with Visual Genome Selected best of two for each sentence according to translation
14CNLP-NITSMMCHMM24en-hi2020/09/18 15:52:353894------33.57---NMTYesPretrained monolingual data (IITB) using Glove and fine-tuned with parallel data and visual features in training using BRNN encoder and doubly-attentive-rnn decoder.
15v036MMCHMM24en-hi2024/08/11 13:55:397328------32.30---NMTNoNMT based system using both image descriptors and text description. A multistage LLM pipeline used for extracting image data descriptions and translation. Fine tuning done in few cases Models Used:
16DCU_NMTMMCHMM24en-hi2024/08/13 03:25:357372------30.30---NMTNoNMT system trained on constrained resources using bert encoded context extracted from visual representation of training data. The context is used only on source side.
17DCU_NMTMMCHMM24en-hi2024/08/11 23:07:087352------28.60---SMTNoContext-aware model that uses image caption data extracted from images as context.
18ORGANIZERMMCHMM24en-hi2020/11/07 01:35:404179------20.34---NMTNo
19ORGANIZERMMCHMM24en-hi2020/11/07 01:39:324178------20.34---NMTNo
20ORGANIZERMMCHMM24en-hi2020/11/07 01:47:444180------20.34---NMTNo
21ODIAGENMMCHMM24en-hi2024/08/15 17:20:587403------ 1.10---OtherNoLLM-based (LLava fine-tuned)
22ODIAGENMMCHMM24en-hi2024/08/15 17:57:507404------ 0.50---OtherNoLLM-based (LLava fine-tuned for 10 epochs)

Notice:

Back to top

RIBES


# Team Task Date/Time DataID RIBES
Method
 Other
Resources
 System
Description
juman kytea mecab moses-
tokenizer		 stanford-
segmenter-
ctb		 stanford-
segmenter-
pku		 indic-
tokenizer		 unuse myseg kmseg
1VoltaMMCHMM24en-hi2021/05/25 13:56:036430------0.859645---NMTYesFinetuned mBART (Used IITB for data augmentation) and added object tags to the input using Mask RCNN
2BITS-PMMCHMM24en-hi2023/07/08 13:44:397124------0.853388---NMTYesNLLB model finetuned on captions + object tags of original & synthetic images using DETR model
300-7MMCHMM24en-hi2024/08/05 15:02:387190------0.842400---NMTYeschallenge set
4ODIAGENMMCHMM24en-hi2023/07/06 03:54:047106------0.815156---NMTNoImage features extracted as Object tags appended with text and MBART fine-tuning
5v036MMCHMM24en-hi2024/08/15 18:19:507406------0.812507---NMTNo
6v036MMCHMM24en-hi2024/08/11 12:43:407319------0.801778---NMTNoNMT based system using both image descriptors and text description. A multistage LLM pipeline used for extracting image data descriptions and translation. Fine tuning done in few cases Models Used:
7v036MMCHMM24en-hi2024/08/11 23:48:237353------0.796730---NMTNo
8239233MMCHMM24en-hi2024/08/13 07:05:337378------0.795538---NMTYesOne-shot prompt for synthetic QA description from captions; translate QA using IndicTrans2; generate caption from QA as context
9v036MMCHMM24en-hi2024/08/14 20:25:457397------0.793972---NMTNo
10CNLP-NITS-PPMMCHMM24en-hi2021/04/27 23:56:015730------0.792097---NMTYesPretrained monolingual data (IITB) using Glove and fine-tuned with parallel data (WAT21 train data+ Extracted Phrase pairs from WAT21 train data +IITB train data) and visual features in training using
11CNLP-NITS-PPMMCHMM24en-hi2022/07/11 12:39:256741------0.791468---NMTNoTransliteration-based phrase pairs augmentation and visual features in training using BRNN encoder and doubly-attentive-rnn decoder.
12iitpMMCHMM24en-hi2021/05/01 23:25:565942------0.790809---NMTNoRemoved special chars at start and end of sentence 1. pre-trained with HindEnCorp, trained with Vis Gen 2. trained with Visual Genome Selected best of two for each sentence according to translation
13SILO_NLPMMCHMM24en-hi2022/07/18 17:33:086959------0.784169---NMTYesObject Tags (Image) + Flickr8 dataset as additional resource + Finetune mBART
14CNLP-NITSMMCHMM24en-hi2020/09/18 15:52:353894------0.754141---NMTYesPretrained monolingual data (IITB) using Glove and fine-tuned with parallel data and visual features in training using BRNN encoder and doubly-attentive-rnn decoder.
15v036MMCHMM24en-hi2024/08/11 13:55:397328------0.752134---NMTNoNMT based system using both image descriptors and text description. A multistage LLM pipeline used for extracting image data descriptions and translation. Fine tuning done in few cases Models Used:
16DCU_NMTMMCHMM24en-hi2024/08/11 23:07:087352------0.711555---SMTNoContext-aware model that uses image caption data extracted from images as context.
17DCU_NMTMMCHMM24en-hi2024/08/13 03:25:357372------0.710342---NMTNoNMT system trained on constrained resources using bert encoded context extracted from visual representation of training data. The context is used only on source side.
18ORGANIZERMMCHMM24en-hi2020/11/07 01:35:404179------0.644230---NMTNo
19ORGANIZERMMCHMM24en-hi2020/11/07 01:39:324178------0.644230---NMTNo
20ORGANIZERMMCHMM24en-hi2020/11/07 01:47:444180------0.644230---NMTNo
21ODIAGENMMCHMM24en-hi2024/08/15 17:20:587403------0.151195---OtherNoLLM-based (LLava fine-tuned)
22ODIAGENMMCHMM24en-hi2024/08/15 17:57:507404------0.116894---OtherNoLLM-based (LLava fine-tuned for 10 epochs)

Notice:

Back to top

AMFM


# Team Task Date/Time DataID AMFM
Method
 Other
Resources
 System
Description
unuse unuse unuse unuse unuse unuse unuse unuse unuse unuse
1VoltaMMCHMM24en-hi2021/05/25 13:56:036430------0.877000---NMTYesFinetuned mBART (Used IITB for data augmentation) and added object tags to the input using Mask RCNN
2iitpMMCHMM24en-hi2021/05/01 23:25:565942------0.823429---NMTNoRemoved special chars at start and end of sentence 1. pre-trained with HindEnCorp, trained with Vis Gen 2. trained with Visual Genome Selected best of two for each sentence according to translation
3CNLP-NITS-PPMMCHMM24en-hi2021/04/27 23:56:015730------0.817356---NMTYesPretrained monolingual data (IITB) using Glove and fine-tuned with parallel data (WAT21 train data+ Extracted Phrase pairs from WAT21 train data +IITB train data) and visual features in training using
4CNLP-NITSMMCHMM24en-hi2020/09/18 15:52:353894------0.787320---NMTYesPretrained monolingual data (IITB) using Glove and fine-tuned with parallel data and visual features in training using BRNN encoder and doubly-attentive-rnn decoder.
5ORGANIZERMMCHMM24en-hi2020/11/07 01:35:404179------0.669760---NMTNo
6ORGANIZERMMCHMM24en-hi2020/11/07 01:39:324178------0.669760---NMTNo
7ORGANIZERMMCHMM24en-hi2020/11/07 01:47:444180------0.669760---NMTNo
8CNLP-NITS-PPMMCHMM24en-hi2022/07/11 12:39:256741------0.000000---NMTNoTransliteration-based phrase pairs augmentation and visual features in training using BRNN encoder and doubly-attentive-rnn decoder.
9SILO_NLPMMCHMM24en-hi2022/07/18 17:33:086959------0.000000---NMTYesObject Tags (Image) + Flickr8 dataset as additional resource + Finetune mBART
10ODIAGENMMCHMM24en-hi2023/07/06 03:54:047106------0.000000---NMTNoImage features extracted as Object tags appended with text and MBART fine-tuning
11BITS-PMMCHMM24en-hi2023/07/08 13:44:397124------0.000000---NMTYesNLLB model finetuned on captions + object tags of original & synthetic images using DETR model
1200-7MMCHMM24en-hi2024/08/05 15:02:387190------0.000000---NMTYeschallenge set
13v036MMCHMM24en-hi2024/08/11 12:43:407319------0.000000---NMTNoNMT based system using both image descriptors and text description. A multistage LLM pipeline used for extracting image data descriptions and translation. Fine tuning done in few cases Models Used:
14v036MMCHMM24en-hi2024/08/11 13:55:397328------0.000000---NMTNoNMT based system using both image descriptors and text description. A multistage LLM pipeline used for extracting image data descriptions and translation. Fine tuning done in few cases Models Used:
15DCU_NMTMMCHMM24en-hi2024/08/11 23:07:087352------0.000000---SMTNoContext-aware model that uses image caption data extracted from images as context.
16v036MMCHMM24en-hi2024/08/11 23:48:237353------0.000000---NMTNo
17DCU_NMTMMCHMM24en-hi2024/08/13 03:25:357372------0.000000---NMTNoNMT system trained on constrained resources using bert encoded context extracted from visual representation of training data. The context is used only on source side.
18239233MMCHMM24en-hi2024/08/13 07:05:337378------0.000000---NMTYesOne-shot prompt for synthetic QA description from captions; translate QA using IndicTrans2; generate caption from QA as context
19v036MMCHMM24en-hi2024/08/14 20:25:457397------0.000000---NMTNo
20ODIAGENMMCHMM24en-hi2024/08/15 17:20:587403------0.000000---OtherNoLLM-based (LLava fine-tuned)
21ODIAGENMMCHMM24en-hi2024/08/15 17:57:507404------0.000000---OtherNoLLM-based (LLava fine-tuned for 10 epochs)
22v036MMCHMM24en-hi2024/08/15 18:19:507406------0.000000---NMTNo

Notice:

Back to top

HUMAN (WAT2022)


# Team Task Date/Time DataID HUMAN
Method
 Other
Resources
System
Description
1CNLP-NITS-PPMMCHMM24en-hi2022/07/11 12:39:256741UnderwayNMTNoTransliteration-based phrase pairs augmentation and visual features in training using BRNN encoder and doubly-attentive-rnn decoder.
2SILO_NLPMMCHMM24en-hi2022/07/18 17:33:086959UnderwayNMTYesObject Tags (Image) + Flickr8 dataset as additional resource + Finetune mBART

Notice:
Back to top

HUMAN (WAT2021)


# Team Task Date/Time DataID HUMAN
Method
 Other
Resources
System
Description
1CNLP-NITS-PPMMCHMM24en-hi2021/04/27 23:56:015730UnderwayNMTYesPretrained monolingual data (IITB) using Glove and fine-tuned with parallel data (WAT21 train data+ Extracted Phrase pairs from WAT21 train data +IITB train data) and visual features in training using
2iitpMMCHMM24en-hi2021/05/01 23:25:565942UnderwayNMTNoRemoved special chars at start and end of sentence 1. pre-trained with HindEnCorp, trained with Vis Gen 2. trained with Visual Genome Selected best of two for each sentence according to translation

Notice:
Back to top

HUMAN (WAT2020)


# Team Task Date/Time DataID HUMAN
Method
 Other
Resources
System
Description
1CNLP-NITSMMCHMM24en-hi2020/09/18 15:52:353894UnderwayNMTYesPretrained monolingual data (IITB) using Glove and fine-tuned with parallel data and visual features in training using BRNN encoder and doubly-attentive-rnn decoder.

Notice:
Back to top

HUMAN (WAT2019)


# Team Task Date/Time DataID HUMAN
Method
 Other
Resources
System
Description

Notice:
Back to top

HUMAN (WAT2018)


# Team Task Date/Time DataID HUMAN
Method
 Other
Resources
System
Description

Notice:
Back to top

HUMAN (WAT2017)


# Team Task Date/Time DataID HUMAN
Method
 Other
Resources
System
Description

Notice:
Back to top

HUMAN (WAT2016)


# Team Task Date/Time DataID HUMAN
Method
 Other
Resources
System
Description

Notice:
Back to top

HUMAN (WAT2015)


# Team Task Date/Time DataID HUMAN
Method
 Other
Resources
System
Description

Notice:
Back to top

HUMAN (WAT2014)


# Team Task Date/Time DataID HUMAN
Method
 Other
Resources
System
Description

Notice:
Back to top

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