Good for Misconceived Reasons: An Empirical Revisiting on the Need for Visual Context in Multimodal Machine Translation
This work challenges the necessity of visual context in MMT, suggesting that prior claims may be misconceived, which could redirect research focus towards interpretability and regularization in multimodal systems.
The study revisits the role of visual context in multimodal machine translation, finding that reported improvements often stem from regularization effects rather than actual use of multimodal information, with models learning to ignore visual inputs while still achieving similar gains.
A neural multimodal machine translation (MMT) system is one that aims to perform better translation by extending conventional text-only translation models with multimodal information. Many recent studies report improvements when equipping their models with the multimodal module, despite the controversy of whether such improvements indeed come from the multimodal part. We revisit the contribution of multimodal information in MMT by devising two interpretable MMT models. To our surprise, although our models replicate similar gains as recently developed multimodal-integrated systems achieved, our models learn to ignore the multimodal information. Upon further investigation, we discover that the improvements achieved by the multimodal models over text-only counterparts are in fact results of the regularization effect. We report empirical findings that highlight the importance of MMT models' interpretability, and discuss how our findings will benefit future research.