统计与规则相结合的藏文人名自动识别研究  

Automatic recognition of tibetan name with the combination of statistics and regular

作　　者：窦嵘 加羊吉[2] 黄伟[3] 

机构地区：[1]甘肃同元信息系统技术有限责任公司 [2]西北民族大学中国民族语言文字信息技术重点实验室,兰州730030 [3]周口师范学院计算机科学技术系,周口466001

出　　处：《长春工程学院学报（自然科学版）》2010年第2期113-115,共3页Journal of Changchun Institute of Technology：Natural Sciences Edition

基　　金：国家自然科学基金(60970071)

摘　　要：提出了一种基于统计与规则相结合的藏文人名自动识别方法。该方法充分挖掘了人名与上下文信息的关联程度,引入互信息对关联程度进行了定量的描述。根据大规模语料(30.2 M文本语料)的统计数据所建立的规则极大地提高了识别算法的准确率;在处理汉族音译人名情况时,采用姓氏驱动的方法,很好地解决了藏文文本中汉族人名的识别问题。实验结果表明,该方法具有较高的准确率和召回率。This paper proposes a recognition method of Tibetan name based on the combination of statistics and regular. The method excavates the association degree of names and the textual information, and then introduces the mutual information to describe the association degree in statistic; statistic the rules which is built from the large-scale corpus （30.2 M text corpus） to improve the accuracy of recognition algorithm; use name-driven method to deal with the pronunciation translating of Han names,in order to get a good solution of recognition problem of Han names in Tibetan text. The experimental result shows that this method has got a higher accuracy and recall rate, and it makes a foundation for future research.

关 键 词：藏文人名识别 互信息 格助词 人名词典 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]