基于N-gram共现的机器翻译自动评估的研究

RESEARCH ON AUTOMATIC EVALUATION OF MT BASED ON N-GRAM CO-OCCURRENCE

机器翻译的研究是人工智能中自然语言理解部分的一个分支,而译文质量是机器翻译系统评估的核心问题.讨论了基于n-gram共现的机器翻译自动评测框架,介绍了BLEU、N IST两种自动评价方法,并针对其提出若干改进思路,旨在更好地实现机器翻译译文质量的自动评估,从而对机器翻译系统的开发和完善提供更好的指导和辅助.

The research of Machine Translation （MT） is one branch of the Natural Language Understanding in the Artificial Intelligence （AI）. At present the quality of the translation has become the key issue in the evaluation. In this paper, the author introduces two automatic evaluation methods BLEU and NIST, based on which suggestions are put forward, so as to manipulate more efficiently the automatic evaluation of MT, and therefore, to guide and improve the development of MT.