Bi-SimCut: A Simple Strategy for Boosting Neural Machine Translation

Anonymous

Abstract: We introduce Bi-SimCut: a simple but effective strategy to boost neural machine translation (NMT) performance. It consists of two training procedures: bidirectional pretraining and unidirectional finetuning. Both procedures utilize SimCut, a simple regularization method that forces the consistency between the output distributions of the original and the cutoff samples. Without utilizing extra dataset via back-translation or integrating large-scale pretrained model, Bi-SimCut achieves strong translation performance across five translation benchmarks (data sizes range from 160K to 20.1M): BLEU scores of $31.16$ for $\texttt{en}\rightarrow\texttt{de}$ and $38.37$ for $\texttt{de}\rightarrow\texttt{en}$ on the IWSLT14 dataset, $30.78$ for $\texttt{en}\rightarrow\texttt{de}$ and $35.15$ for $\texttt{de}\rightarrow\texttt{en}$ on the WMT14 dataset, and $27.17$ for $\texttt{zh}\rightarrow\texttt{en}$ on the WMT17 dataset. SimCut is not a new method, but a version of Cutoff (Shen et al., 2020) simplified and adapted for NMT, and it could be considered as a perturbation-based method. Given the universality and simplicity of Bi-SimCut and SimCut, we believe they can serve as strong baselines for future NMT research.

Paper Type: long