摘要
针对专利文献翻译中的复杂并列结构的处理,提出了一种预处理和后处理的方法——拆分合并法。它弥补了统计翻译系统在复杂结构方面的劣势,同时保留统计翻译在经验句式、人力成本等方面优于规则系统的特点。实验表明,这种方法稳定地提高了翻译系统的准确率。此外,它独立于具体的统计翻译系统,可以方便地移植到不同的统计翻译系统上。
In order to process parallel structures during translating patent articles,a pre-proeessing and post-processing method （Separating-combining Method） is proposed.It fixes disadvantages of the statistical translation system in processing complex structures,while maintains advantages in processing idioms and saving cost of human labor.Experiments show that this method consistently improves the accuracy of translation system.Besides,it is independent of the particularity of statistical translation system,and can be conveniently applied on different systems.
出处
《计算机工程与应用》
CSCD
北大核心
2009年第24期125-129,共5页
Computer Engineering and Applications
关键词
机器翻译
语法结构
专利文献
machine translation
syntactic structure
patent article
