You Are What You Train: Effects of Data Composition on Training Context-aware Machine Translation Models
This work addresses the challenge of achieving coherent translations in machine translation, particularly for pronoun disambiguation, but it is incremental as it builds on existing hypotheses about data sparsity.
The study tackled the problem of context utilization in machine translation by systematically validating that sparsity of contextually rich examples in training data is a key bottleneck, and proposed training strategies that improved context utilization with accuracy gains of up to 6 and 8 percentage points on evaluation metrics.
Achieving human-level translations requires leveraging context to ensure coherence and handle complex phenomena like pronoun disambiguation. Sparsity of contextually rich examples in the standard training data has been hypothesized as the reason for the difficulty of context utilization. In this work, we systematically validate this claim in both single- and multilingual settings by constructing training datasets with a controlled proportions of contextually relevant examples. We demonstrate a strong association between training data sparsity and model performance confirming sparsity as a key bottleneck. Importantly, we reveal that improvements in one contextual phenomenon do no generalize to others. While we observe some cross-lingual transfer, it is not significantly higher between languages within the same sub-family. Finally, we propose and empirically evaluate two training strategies designed to leverage the available data. These strategies improve context utilization, resulting in accuracy gains of up to 6 and 8 percentage points on the ctxPro evaluation in single- and multilingual settings respectively.