Zero-Resource Translation with Multi-Lingual Neural Machine Translation
Orhan Firat⋆ Middle East Technical University、Baskaran Sankaran IBM T.J. Watson Research Center \ANDYaser Al-Onaizan IBM T.J. Watson Research Center、Fatos T. Yarman Vural Middle East Technical University、Kyunghyun Cho New York University
Abstract
In this paper, we propose a novel finetuning algorithm for the recently introduced multi-way, mulitlingual neural machine translate that enables zero-resource machine translation. When used together with novel many-to-one translation strategies, we empirically show that this finetuning algorithm allows the multi-way, multilingual model to translate a zero-resource language pair (1) as well as a single-pair neural translation model trained with up to 1M direct parallel sentences of the same language pair and (2) better than pivot-based translation strategy, while keeping only one additional copy of attention-related parameters.
中文速览
多语言神经机器翻译模型通常无法直接翻译训练时从未见过直接平行语料的语言对，即"零资源翻译"（zero-resource translation）问题。本文针对一种多路、多语言神经翻译模型提出了一种新型微调算法：利用"多对一"翻译策略生成伪平行语料，再仅微调注意力机制相关参数（额外增加一份注意力参数副本），从而绕开目标语言对直接平行数据的需求。在西班牙语、法语、英语的大规模实验中，该方法使零资源翻译效果达到了用多达100万句真实平行句对训练的单语言对模型的水平，同时也超越了传统的枢轴语言（pivot-based）翻译方案。这一结果表明，多语言神经翻译模型在低资源乃至零资源语言翻译场景中具有巨大潜力，且额外引入的参数开销极小。
原文 arXiv:1606.04164;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1606.04164v1