结合多特征的支持向量机中文组织机构名识别模型

Fusion of Multiple Features for SVM Chinese Organization Names Recognization Model

以支持向量机(SVM)为基本框架,提出一种结合多特征的支持向量机中文组织机构名识别模型。考虑中文组织机构名的特点,抽取局部特征与全局特征,并将特征向量转化为二进制表示,在此基础上建立训练集。基于1998年《人民日报》语料的实验结果表明,该混合模型对中文组织机构名的识别是有效的。同时基于不同测试数据的实验结果表明,该模型对不同测试数据源具有一致性。

Proposes a hybrid pattern for SVM Chinese Organization names, which fuses multiple features. Thinking about features of Chinese organization names, abstracts local features and global fea- tures, and expresses features-vector in binary, and establishes the training collection. From the experimental results on testing set for People＇s Daily in 1998, it can be concluded that the es- tablished hybrid model is effective on recognization for chinese organization names. And the ex- periments on another different testing set also confirm the above conclusion, which shows that this algorithm has consistence on different testing data sources.