Google’s Multilingual Neural Machine Translation System: Enabling Zero-Shot Translation

Abstract

在多语言翻译中,本文提出解决方案为不修改标准的NMT系统,而是在开头加入人工的token去特定标记需要的目标语言。本工作比先前的Multilingual NMT要简单的多,并且不需要加入任何的参数。

Introduce

这样的学习有几大优势:

  • Simplicity
  • Low-resource language improvement
  • Zero-shot translation

Related Work

  • zero-shot translation
  • zero-resource : fine-tune

System Architecture for Multilingual Translation

  • introduce an artificial token :

original : Hello, how are you? -> Hola, ¿cómo estás?

result: <2es> Hello, how are you? -> Hola, ¿cómo estás?

你可能感兴趣的:(Google’s Multilingual Neural Machine Translation System: Enabling Zero-Shot Translation)