English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Multimodal machine translation (MMT) systems have been shown to outperformtheir text-only neural machine translation (NMT) counterparts when visualcontext is available. However, recent studies have also shown that theperformance of MMT models is only marginally impacted when the associated imageis replaced with an unrelated image or noise, which suggests that the visualcontext might not be exploited by the model at all. We hypothesize that thismight be caused by the nature of the commonly used evaluation benchmark, alsoknown as Multi30K, where the translations of image captions were preparedwithout actually showing the images to human translators. In this paper, wepresent a qualitative study that examines the role of datasets in stimulatingthe leverage of visual modality and we propose methods to highlight theimportance of visual signals in the datasets which demonstrate improvements inreliance of models on the source images. Our findings suggest the research oneffective MMT architectures is currently impaired by the lack of suitabledatasets and careful consideration must be taken in creation of future MMTdatasets, for which we also provide useful insights.

Vision Matters When It Should: Sanity Checking Multimodal Machine Translation Models