Multilingual machine translation (MMT), trained on a mixture of parallel and monolingual data, is key for improving translation in low-resource language pairs. However, the literature offers conflicting results on the performance of different methods of including monolingual data. To resolve this, we examine how denoising autoencoding (DAE) and backtranslation (BT) impact MMT under different data conditions and model scales. Unlike prior studies, we use a realistic dataset of 100 translation directions and consider many domain combinations of monolingual and test data. We find that monolingual data generally helps MMT, but models are surprisingly brittle to domain mismatches, especially at smaller model scales. BT is beneficial when the parallel, monolingual, and test data sources are similar but can be detrimental otherwise, while DAE is less effective than previously reported. Next, we analyze the impact of scale (from 90M to 1.6B parameters) and find it is important for both methods, particularly DAE. As scale increases, DAE transitions from underperforming the parallel-only baseline at 90M to converging with BT performance at 1.6B, and even surpassing it in low-resource. These results offer new insights into how to best use monolingual data in MMT.

翻译：多语言机器翻译（MMT）通过混合使用平行数据和单语数据进行训练，对于改善低资源语言对的翻译质量至关重要。然而，关于不同单语数据纳入方法的性能，文献中呈现了相互矛盾的结果。为解决这一问题，我们研究了去噪自编码（DAE）和回译（BT）在不同数据条件和模型规模下对MMT的影响。与以往研究不同，我们使用了包含100个翻译方向的真实数据集，并考虑了单语数据与测试数据的多种领域组合。我们发现单语数据通常有助于MMT，但模型对领域不匹配表现出惊人的脆弱性，尤其在较小模型规模下。当平行数据、单语数据和测试数据来源相似时，BT表现有益，否则可能产生负面影响；而DAE的效果则低于此前报道水平。此外，我们分析了模型规模（从90M到1.6B参数）的影响，发现规模对两种方法均至关重要，尤其对DAE。随着规模增大，DAE从90M参数时表现不如仅使用平行数据的基线模型，逐步演进到1.6B参数时与BT性能趋同，甚至在低资源场景下超越BT。这些结果为如何在MMT中最优地使用单语数据提供了新洞见。