LIUM-CVC Submissions for WMT17 Multimodal Translation Task
About
This paper describes the monomodal and multimodal Neural Machine Translation systems developed by LIUM and CVC for WMT17 Shared Task on Multimodal Translation. We mainly explored two multimodal architectures where either global visual features or convolutional feature maps are integrated in order to benefit from visual context. Our final systems ranked first for both En-De and En-Fr language pairs according to the automatic evaluation metrics METEOR and BLEU.
Ozan Caglayan, Walid Aransa, Adrien Bardet, Mercedes Garc\'ia-Mart\'inez, Fethi Bougares, Lo\"ic Barrault, Marc Masana, Luis Herranz, Joost van de Weijer• 2017
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Multimodal Machine Translation | Multi30K (test) | BLEU-454.7 | 139 | |
| Multimodal Machine Translation (English-German) | Multi30K 2016 (test) | BLEU37.8 | 52 | |
| Multimodal Machine Translation | Multi30k En-De 2017 (test) | METEOR52.2 | 45 | |
| Multimodal Machine Translation | Multi30k En-Fr 2017 (test) | METEOR69.5 | 31 | |
| Multimodal Machine Translation | Multi30k En-Fr 2016 (test) | METEOR Score71.3 | 30 | |
| Multimodal Machine Translation | MSCOCO EN-FR (test) | BLEU43.5 | 19 | |
| Multi-modal Machine Translation | Multi30k WMT17 (test) | BLEU30.7 | 16 | |
| Multimodal Machine Translation | MSCOCO Ambiguous EN-DE (test) | BLEU26.4 | 13 | |
| Multimodal Machine Translation (English to German) | Ambiguous COCO WMT2017 (test) | BLEU27.1 | 11 | |
| Machine Translation | Ambiguous COCO | BLEU26.4 | 6 |
Showing 10 of 15 rows