Lexically Cohesive Neural Machine Translation with Copy Mechanism

TL;DR

This paper introduces a copy mechanism in neural machine translation to explicitly enhance lexical cohesion across document translations, leading to more consistent word choices.

Contribution

It presents a novel explicit approach for lexical cohesion in neural translation models by integrating a copy mechanism, improving over previous implicit methods.

Findings

Significant improvement in lexical cohesion for Japanese-English translation

Model outperforms previous context-aware neural translation models

Demonstrates effectiveness of explicit copying for discourse consistency

Abstract

Lexically cohesive translations preserve consistency in word choices in document-level translation. We employ a copy mechanism into a context-aware neural machine translation model to allow copying words from previous translation outputs. Different from previous context-aware neural machine translation models that handle all the discourse phenomena implicitly, our model explicitly addresses the lexical cohesion problem by boosting the probabilities to output words consistently. We conduct experiments on Japanese to English translation using an evaluation dataset for discourse translation. The results showed that the proposed model significantly improved lexical cohesion compared to previous context-aware models.