维吾尔语数词类命名实体的识别与翻译  

RECOGNITION AND TRANSLATION OF UYGHUR NAMED ENTITIES IN NUMERALS CLASS

作　　者：张磊[1,2] 杨雅婷[1] 米成刚[1,2] 李晓[1] 

机构地区：[1]中国科学院新疆理化技术研究所,新疆乌鲁木齐830011 [2]中国科学院大学,北京100049

出　　处：《计算机应用与软件》2015年第8期64-67,109,共5页Computer Applications and Software

基　　金：中国科学院战略性先导科技专项项目(XDA06030400);中国科学院"西部之光"人才培养计划"西部博士"项目(XBBS201216);新疆维吾尔自治区青年科技创新人才培养工程项目(2013731021);中国科学院西部行动计划项目(KGZD-EW-501)

摘　　要：针对维吾尔语数词类命名实体(时间、日期、货币、百分比)在维汉机器翻译中翻译不准确的问题,分析其构成规律及边界信息,设计基于维汉平行语料的维吾尔语数词类命名实体的识别与翻译系统。通过有限自动机结合触发词识别并翻译维语基本数词,从平行语料中自动抽取出翻译模板,匹配模板并实现翻译。实验表明,维吾尔语数词类命名实体的识别F值达到了91%,有效提高了维汉机器翻译的质量。Aiming at the problem that Uyghur named entities（ time,date,money,percentage） in numerals class are inaccurately translated in Uyghur-Chinese machine translation,we designed a Uyghur-Chinese parallel corpus-based recognition and translation system for Uyghur named entities in numerals class by analysing the formation laws and boundary information of these named entities. Uyghur basic numerals are recognised and translated through finite automata in combination with triggering words,and the translation templates will be automatically extracted from Uyghur-Chinese parallel corpus,the templates will then be matched to implement the translation. The F value of recognition achieves 91% in Uyghur named entities in numerals class,the system effectively improves the quality of Uyghur-Chinese machine translation.

关 键 词：平行语料 数词类 命名实体 维汉机器翻译 有限自动机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]