In this paper, we look at the case of a Generic text-to-text NMT model that has to deal with data coming from various modalities, like speech, images, or noisy text extracted from the web. We propose a two-step method, based on composable adapters, to deal with this problem of Multimodal Robustness. In a first step, we separately learn domain adapters and modality specific adapters, to deal with noisy input coming from various sources: ASR, OCR, or noisy text (UGC). In a second step, we combine these components at runtime via dynamic routing or, when the source of noise is unknown, via two new transfer learning mechanisms (Fast Fusion and Multi Fusion). We show that our method provides a flexible, state-of-the-art, architecture able to deal with noisy multimodal inputs.

We introduce our TMEKU system submitted to the English-Japanese Multimodal Translation Task for WAT 2021. We participated in the Flickr30kEnt-JP task and Ambiguous MSCOCO Multimodal task under the constrained condition using only the officially provided datasets. Our proposed system employs soft alignment of word-region for multimodal neural machine translation (MNMT). The experimental results evaluated on the BLEU metric provided by the WAT 2021 evaluation site show that the TMEKU system has achieved the best performance among all the participated systems. Further analysis of the case study demonstrates that leveraging word-region alignment between the textual and visual modalities is the key to performance enhancement in our TMEKU system, which leads to better visual information use.

Existing studies on multimodal neural machine translation (MNMT) have mainly focused on the effect of combining visual and textual modalities to improve translations. However, it has been suggested that the visual modality is only marginally beneficial. Conventional visual attention mechanisms have been used to select the visual features from equally-sized grids generated by convolutional neural networks (CNNs), and may have had modest effects on aligning the visual concepts associated with textual objects, because the grid visual features do not capture semantic information. In contrast, we propose the application of semantic image regions for MNMT by integrating visual and textual features using two individual attention mechanisms (double attention). We conducted experiments on the Multi30k dataset and achieved an improvement of 0.5 and 0.9 BLEU points for English-German and English-French translation tasks, compared with the MNMT with grid visual features. We also demonstrated concrete improvements on translation performance benefited from semantic image regions.