基于短语模板对齐的统计机器翻译系统  

The Phrase-Template Alignment Based Statistical Machine Translation

作　　者：杨振东[1] 庞薇[1] 魏玮[1] 杜金华[1] 陈振标[1] 宗成庆[1] 

机构地区：[1]中国科学院自动化研究所高技术创新中心,北京100080

出　　处：《中文信息学报》2006年第B03期53-60,共8页Journal of Chinese Information Processing

基　　金：国家自然科学基金资助项目（60272041）

摘　　要：论文提出了一种基于短语模板时齐的机器翻译系统。系统采用基于短语模板对齐的翻译模型替代原始基于词的模型，提出了计算短语模板翻译概率的方法，改进了先前单纯基于短语的搜索算法和回溯方法。解码时引入繁衍度为0且出现频率高的词，使翻译结果更合理。论文进行了一系列实验，介绍了05年863评洲，证明本系统翻译结果的Bleu得分比原来的系统有了显著提高。In this paper we propose a phrase-template alignment based translation system, in the system, we use phrase-template translation model instead of word-hased model. An improved method to compute phrase-template translation probability is studied. A phrase-template based decoder we developed employs a beam search algorithm, in which some target language words that have both high frequency of appearance and also fertility zero are introduced to make the result more reasonable, We improve the previously proposed tracing back algorithm to get the best path. Some experiments concernod are presented.

关 键 词：统计机器翻译 翻译模型 搜索算法 

分 类 号：TP391.2[自动化与计算机技术—计算机应用技术]