Debiasing Word Embeddings Improves Multimodal Machine Translation

TL;DR

This paper investigates the impact of debiasing pretrained word embeddings on multimodal neural machine translation, showing that appropriate debiasing techniques can significantly improve translation quality.

Contribution

It introduces two debiasing methods for pretrained embeddings and demonstrates their effectiveness across multiple models and language pairs in multimodal NMT.

Findings

Performance improved by up to +1.93 BLEU and +2.02 METEOR for English-German.

Debiasing techniques enhance translation quality in multimodal NMT.

Effective debiasing reduces hubness in high-dimensional embeddings.

Abstract

In recent years, pretrained word embeddings have proved useful for multimodal neural machine translation (NMT) models to address the shortage of available datasets. However, the integration of pretrained word embeddings has not yet been explored extensively. Further, pretrained word embeddings in high dimensional spaces have been reported to suffer from the hubness problem. Although some debiasing techniques have been proposed to address this problem for other natural language processing tasks, they have seldom been studied for multimodal NMT models. In this study, we examine various kinds of word embeddings and introduce two debiasing techniques for three multimodal NMT models and two language pairs -- English-German translation and English-French translation. With our optimal settings, the overall performance of multimodal models was improved by up to +1.93 BLEU and +2.02 METEOR for English-German translation and +1.73 BLEU and +0.95 METEOR for English-French translation.