Rethinking Zero-shot Neural Machine Translation: From a Perspective of Latent Variables

TL;DR

This paper proposes a denoising autoencoder approach with pivot language to enhance zero-shot translation in multilingual NMT, effectively reducing spurious correlations and outperforming existing methods.

Contribution

It introduces a novel training objective that implicitly maximizes zero-shot translation probabilities, improving transfer performance in multilingual NMT.

Findings

Significantly outperforms state-of-the-art methods on benchmark datasets.

Effectively eliminates spurious correlations in zero-shot translation.

Theoretically shows implicit maximization of zero-shot translation probabilities.

Abstract

Zero-shot translation, directly translating between language pairs unseen in training, is a promising capability of multilingual neural machine translation (NMT). However, it usually suffers from capturing spurious correlations between the output language and language invariant semantics due to the maximum likelihood training objective, leading to poor transfer performance on zero-shot translation. In this paper, we introduce a denoising autoencoder objective based on pivot language into traditional training objective to improve the translation accuracy on zero-shot directions. The theoretical analysis from the perspective of latent variables shows that our approach actually implicitly maximizes the probability distributions for zero-shot directions. On two benchmark machine translation datasets, we demonstrate that the proposed method is able to effectively eliminate the spurious correlations and significantly outperforms state-of-the-art methods with a remarkable performance. Our code is available at https://github.com/Victorwz/zs-nmt-dae.