Learning to Parse and Translate Improves Neural Machine Translation

TL;DR

This paper introduces NMT+RNNG, a hybrid model that integrates linguistic parsing with neural machine translation, improving translation quality by incorporating linguistic priors during training.

Contribution

It presents a novel hybrid model combining RNNG with NMT, enabling better linguistic integration and translation performance.

Findings

Effective across four language pairs

Improves translation quality with linguistic priors

Encourages linguistic-aware translation during training

Abstract

There has been relatively little attention to incorporating linguistic prior to neural machine translation. Much of the previous work was further constrained to considering linguistic prior on the source side. In this paper, we propose a hybrid model, called NMT+RNNG, that learns to parse and translate by combining the recurrent neural network grammar into the attention-based neural machine translation. Our approach encourages the neural machine translation model to incorporate linguistic prior during training, and lets it translate on its own afterward. Extensive experiments with four language pairs show the effectiveness of the proposed NMT+RNNG.