基于枢轴语言的多语言神经机器翻译研究

Research on Multilingual Neural Machine Translation based on Pivot Language

机器翻译多年来的发展,虽取得显著进步,但是离完美取代人工翻译还有很长的路要走。不仅需要克服自身方法或算法的不足,还受到平行语料规模和质量的制约。本文提出用枢轴语言作为桥梁,借助已有的机器翻译技术,通过构建伪平行语料以及借助中间语言实现多语言机器翻译,并取得很好的翻译效果。

Machine translation has made remarkable progress over the years, however, it still has a long way to go to replace human translation perfectly. It not only need overcome the shortcomings of its own methods or algorithms, but also is restricted by the size and quality of parallel corpus. In this paper, we use pivot language as a bridge with the aid of existing machine translation technology and realize multilingual machine translation by constructing pseudo parallel corpus and intermediate language. Finally, we achieved good translation results.