基于短语的统计机器翻译中短语抽取算法改进  

Improving phrase-based statistical translation by modifying phrase extraction algorithm

作　　者：强静[1] 张建[1] 

机构地区：[1]中国科学院合肥智能机械研究所

出　　处：《计算机工程与应用》2008年第13期147-149,共3页Computer Engineering and Applications

基　　金：中国科学院知识创新工程重要方向项目(No.KGCX2-SW-511)

摘　　要：针对基于短语统计机器翻译中目前常用的Och提出的短语抽取算法,提出了一种改进算法。该算法能够在原有算法的基础上抽取出更多的准确对齐信息,这对语料库较小的汉民统计机器来说意义重大,增加正确的对齐信息可以减少未登录词的产生,提高翻译正确率。经过对不同规模语料库的实验,抽取的短语对数目有明显增多。The paper proposes an improved algorithm of Phrase Extract based on the Och＇s phrase extraction algorithm in the phrase based statistical machine translation.The algorithm can take more accurate alignment information based on the original algorithm.It is of great significance for the smaller corpus statistical machinery.It can reduce the unknown words by increasing in correct alignment information,and increases the rate of correct translation.After the different scale corpus experiment.The extracted number of phrase is obviously increase.

关 键 词：统计机器翻译 翻译模型 短语抽取 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]