Faster Transformer Decoding: N-gram Masked Self-Attention

Motivated by the fact that most of the information relevant to the prediction of target tokens is drawn from the source sentence S=s1, ldots, sS, we propose truncating the target-side window used for computing self-attention by making an N-gram assumption. Experiments on WMT EnDe and EnFr data sets show that the N-gram masked self-attention model loses very little in BLEU score for N values in the range 4, ldots, 8, depending on the task.