基于HMM的藏语语料库词性自动标注研究  

Researches of Part of Speech Tagging for Tibetan Texts Based on Hidden Markov Model

作　　者：苏俊峰[1] 祁坤钰[1] 本太[1] 

机构地区：[1]西北民族大学中国民族信息技术研究院,甘肃兰州730030

出　　处：《西北民族大学学报（自然科学版）》2009年第1期42-45,共4页Journal of Northwest Minzu University(Natural Science)

基　　金：<现代藏语语法信息词典>建设研究;甘肃省社科规划项目;项目编号2007-04

摘　　要：针对藏语区别于英语和汉语,分析藏语的构形特征,得到词性标注集.从人工标注的语料中统计词和词性频率以及训练得到二元语法的HMM模型参数,运用Viterbi算法完成基于统计方法的词性标注.After the characteristics of word formation in Tibetan which are different from English and Chinese are analyzed, tagging set on part of speech is obtained. Then from the man - made - corpus, the appearance frequency of words and part of speech frequency are worked out, and HMM＇ s module parameter of binary grammar is acquired. Through the Viterbi Algorithm, tagging part of speech which is based on the method of statistics is finished. Finally, the tagging is corrected with the application of Tibetan rules.

关 键 词：藏语词性标注 词性标注集 HMM 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]