Correcting Length Bias in Neural Machine Translation
This addresses length bias in neural machine translation, which is an incremental improvement for translation systems.
The paper tackled the problems of beam search hurting translation quality and neural machine translation producing overly short outputs, showing that correcting the brevity issue nearly eliminates the beam problem and that a simple per-word reward tuned with the perceptron algorithm works effectively.
We study two problems in neural machine translation (NMT). First, in beam search, whereas a wider beam should in principle help translation, it often hurts NMT. Second, NMT has a tendency to produce translations that are too short. Here, we argue that these problems are closely related and both rooted in label bias. We show that correcting the brevity problem almost eliminates the beam problem; we compare some commonly-used methods for doing this, finding that a simple per-word reward works well; and we introduce a simple and quick way to tune this reward using the perceptron algorithm.