Multi-Source Neural Translation

TL;DR

This paper introduces a multi-source neural machine translation model that leverages French and German sources to improve English translation quality, achieving significant BLEU score improvements over existing models.

Contribution

It presents a novel multi-source neural translation framework and explores various combination methods, resulting in up to +4.8 BLEU improvements.

Findings

Achieved up to +4.8 BLEU score increase

Demonstrated effectiveness of multi-source approach

Explored multiple combination strategies

Abstract

We build a multi-source machine translation model and train it to maximize the probability of a target English string given French and German sources. Using the neural encoder-decoder framework, we explore several combination methods and report up to +4.8 Bleu increases on top of a very strong attention-based neural translation model.