基于单个词语特征模板的汉语词性标注  

Chinese Part-of-speech Tagging Based on Single Word Feature Templates

作　　者：于江德[1] 周宏宇[1] 余正涛[2] 

机构地区：[1]安阳师范学院计算机与信息工程学院,河南安阳455002 [2]昆明理工大学信息工程与自动化学院,云南昆明650051

出　　处：《山西大学学报（自然科学版）》2011年第4期513-517,共5页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(60663004);河南省高等学校青年骨干教师项目(2009GGJS-108)

摘　　要：针对实际应用中语言模型应该占用更小存储空间且加载速度快等需求,采用最大熵模型进一步研究了汉语词性标注中设定的特征模板集和训练后模型大小、标注精度等指标之间的关系,并在国际汉语分词评测Bake-off2007的PKU、NCC、CTB三种语料上进行了对比实验.实验结果表明,双词语组合特征模板大大增加了训练后模型的大小,对汉语词性标注精度却没有提高,而基于单个词语特征模板训练后的模型大小不足原先大小的1/5,标注精度却没有下降.Language model for practical application should take up less storage space and loading speed,so we studied the relations of the feature template set and the training model size,tagging accuracy for Chinese part-of-speech tagging by using maximum entropy model.The closed evaluations are performed on PKU,NCC and CTB corpus from the Bakeoff-2007,and the comparative experiments are performed on different feature templates.The experimental results show:the size of the model has greatly increased and tagging accuracy is not improved using combination of two-word feature templates,while the size of the model is 1/5 of the original size,but without any marked decline in accuracy using the single word feature template.

关 键 词：汉语词性标注 单个词语特征模板 最大熵模型 上下文 上下文窗口 

分 类 号：N55[自然科学总论]