Neural Machine Translation with Reconstruction

TL;DR

This paper introduces an encoder-decoder-reconstructor framework for neural machine translation that enhances translation adequacy by reconstructing source sentences from target representations, leading to improved translation quality.

Contribution

It proposes a novel reconstruction-based framework for NMT that significantly improves translation adequacy compared to existing models.

Findings

Significant improvement in translation adequacy.

Outperforms state-of-the-art NMT and statistical MT systems.

Reconstruction mechanism effectively captures source information.

Abstract

Although end-to-end Neural Machine Translation (NMT) has achieved remarkable progress in the past two years, it suffers from a major drawback: translations generated by NMT systems often lack of adequacy. It has been widely observed that NMT tends to repeatedly translate some source words while mistakenly ignoring other words. To alleviate this problem, we propose a novel encoder-decoder-reconstructor framework for NMT. The reconstructor, incorporated into the NMT model, manages to reconstruct the input source sentence from the hidden layer of the output target sentence, to ensure that the information in the source side is transformed to the target side as much as possible. Experiments show that the proposed framework significantly improves the adequacy of NMT output and achieves superior translation result over state-of-the-art NMT and statistical MT systems.