Poor Man's Quality Estimation: Predicting Reference-Based MT Metrics Without the Reference
This work addresses the computational and annotation cost limitations in machine translation quality estimation for researchers and practitioners, though it is incremental as it builds on existing metric estimation ideas.
The paper tackles the problem of predicting automated machine translation metrics without access to the reference translation, achieving correlations of up to 60% for BLEU and 51% for other metrics, and shows that pre-training on this task improves quality estimation performance from 20% to 23% correlation.
Machine translation quality estimation (QE) predicts human judgements of a translation hypothesis without seeing the reference. State-of-the-art QE systems based on pretrained language models have been achieving remarkable correlations with human judgements yet they are computationally heavy and require human annotations, which are slow and expensive to create. To address these limitations, we define the problem of metric estimation (ME) where one predicts the automated metric scores also without the reference. We show that even without access to the reference, our model can estimate automated metrics ($ρ$=60% for BLEU, $ρ$=51% for other metrics) at the sentence-level. Because automated metrics correlate with human judgements, we can leverage the ME task for pre-training a QE model. For the QE task, we find that pre-training on TER is better ($ρ$=23%) than training for scratch ($ρ$=20%).