句对齐研究综述

A Survey of Sentence Alignment

神经机器翻译是目前机器翻译领域的主流方法,拥有足够数量的双语平行语料是训练出一个好的翻译模型的前提。双语句对齐技术作为一种从不同语言端单语语料中获取双语平行句对的技术,因此得到广泛的研究。该文首先简单介绍句对齐任务及其相应的评测标准,然后归纳总结前人在句对齐任务上的研究进展,以及句对齐任务的相关信息,并简单概括参加团队所提交的系统,最后对当前工作进行总结并展望未来的工作。

Neural machine translation is currently the mainstream in the field of machine translation,while a large amount of parallel corpora is necessary to train a good neural machine translation system.As a task to explore methods to get parallel sentence from a mass of candidate monolingual text,sentence alignment technology has been widely studied.This paper briefly introduces the sentence alignment task and it’s evaluation criterion.Then it summarizes the previous research in the field of sentence alignment.The shared tasks on sentence alignment and the system submitted by the participating teams are also discussed.Finally we summarize the current work on sentence alignment and give an outlook for its future development.