Greedy Search with Probabilistic N-gram Matching for Neural Machine Translation

TL;DR

This paper introduces a differentiable sequence-level training method for neural machine translation that uses probabilistic n-gram matching and greedy search, improving translation quality and stability over reinforcement-based approaches.

Contribution

It proposes a novel differentiable training objective based on probabilistic n-gram matching that avoids reinforcement learning and employs greedy search during training to reduce exposure bias.

Findings

Achieves 1.5 BLEU point improvement over strong baseline

Outperforms reinforcement-based algorithms significantly

Demonstrates stable and effective training for NMT

Abstract

Neural machine translation (NMT) models are usually trained with the word-level loss using the teacher forcing algorithm, which not only evaluates the translation improperly but also suffers from exposure bias. Sequence-level training under the reinforcement framework can mitigate the problems of the word-level loss, but its performance is unstable due to the high variance of the gradient estimation. On these grounds, we present a method with a differentiable sequence-level training objective based on probabilistic n-gram matching which can avoid the reinforcement framework. In addition, this method performs greedy search in the training which uses the predicted words as context just as at inference to alleviate the problem of exposure bias. Experiment results on the NIST Chinese-to-English translation tasks show that our method significantly outperforms the reinforcement-based algorithms and achieves an improvement of 1.5 BLEU points on average over a strong baseline system.