Unsupervised Multimodal Neural Machine Translation with Pseudo Visual Pivoting

TL;DR

This paper introduces an unsupervised multimodal neural machine translation model that leverages visual content and pseudo visual pivoting to improve language translation accuracy without requiring images during testing.

Contribution

It proposes a novel approach combining multimodal back-translation and pseudo visual pivoting to enhance unsupervised multimodal translation in a shared visual-semantic space.

Findings

Significant improvement over state-of-the-art methods on Multi30K dataset.

Effective even when images are not available during testing.

Demonstrates the potential of visual content in unsupervised translation.

Abstract

Unsupervised machine translation (MT) has recently achieved impressive results with monolingual corpora only. However, it is still challenging to associate source-target sentences in the latent space. As people speak different languages biologically share similar visual systems, the potential of achieving better alignment through visual content is promising yet under-explored in unsupervised multimodal MT (MMT). In this paper, we investigate how to utilize visual content for disambiguation and promoting latent space alignment in unsupervised MMT. Our model employs multimodal back-translation and features pseudo visual pivoting in which we learn a shared multilingual visual-semantic embedding space and incorporate visually-pivoted captioning as additional weak supervision. The experimental results on the widely used Multi30K dataset show that the proposed model significantly improves over the state-of-the-art methods and generalizes well when the images are not available at the testing time.