基于统计机器翻译的汉维词对齐研究  

RESEARCH ON STATISTICAL MACHINE TRANSLATION-BASED CHINESE-UYGHUR WORD ALIGNMENT

作　　者：刘建明[1] 吐尔根.依不拉音 艾山.吾买尔[1] 

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046

出　　处：《计算机应用与软件》2011年第4期57-59,90,共4页Computer Applications and Software

基　　金：国家自然科学基金(60663006);国家语委科研项目(MZ115-75)

摘　　要：描述了一个基于统计机器翻译的汉维词对齐系统。系统处理过程分为两个模块:预处理和词对齐。预处理过程包括汉文文本预处理和维吾尔文文本预处理,其中维吾尔文文本预处理过程为:首先将维吾尔文转换成拉丁维文,然后将拉丁维文中个别字符替换为无歧义的字符。词对齐实现过程:首先利用IBM Model1-3,然后结合Och等人提出的启发式的思路进行优化,构建基于统计机器翻译的汉维词对齐系统。实验结果表明此系统可行。This paper describes a Chinese-Uyghur word alignment system which is based on statistical machine translation.There are two models in processing procedure of the system：pre-process and word alignment.The pre-process includes Chinese text pre-process and Uyghur text pre-process,in it the pre-process procedure for Uyghur text is as follows：First the Uyghur is transferred to Latin-Uyghur,then the exceptional characters in Latin-Uyghur will be replaced by the unambiguous characters.The implementation process of word alignment is：First,IBM model 1-3 is used,then in combination with the heuristic theory by Och the optimisation is conducted,and a Chinese-Uyghur word alignment system based on statistical machine translation is constructed.Experimental results show that this system is feasible.

关 键 词：词对齐 IBM Model1-3 启发式优化 

分 类 号：TP391.2[自动化与计算机技术—计算机应用技术]