Rethinking the Reasonability of the Test Set for Simultaneous Machine Translation
This addresses evaluation challenges for researchers and practitioners in simultaneous machine translation, but it is incremental as it focuses on dataset refinement rather than a new model.
The paper tackles the problem that standard test sets underestimate simultaneous machine translation (SimulMT) performance because they are not designed for monotonic alignment, and it shows that using a manually annotated monotonic test set (SiMuST-C) alleviates this issue and finetuning on a monotonic training set improves models by up to 3 BLEU points.
Simultaneous machine translation (SimulMT) models start translation before the end of the source sentence, making the translation monotonically aligned with the source sentence. However, the general full-sentence translation test set is acquired by offline translation of the entire source sentence, which is not designed for SimulMT evaluation, making us rethink whether this will underestimate the performance of SimulMT models. In this paper, we manually annotate a monotonic test set based on the MuST-C English-Chinese test set, denoted as SiMuST-C. Our human evaluation confirms the acceptability of our annotated test set. Evaluations on three different SimulMT models verify that the underestimation problem can be alleviated on our test set. Further experiments show that finetuning on an automatically extracted monotonic training set improves SimulMT models by up to 3 BLEU points.