基于类向量模型的中文姓名识别研究  

Research of Chinese Names Identification Based on Class Vector Model

作　　者：贾品贵[1] 杨一平[1] 卢朋[1] 

机构地区：[1]中国科学院自动化所综合信息中心,北京100080

出　　处：《计算机应用研究》2007年第4期111-113,共3页Application Research of Computers

摘　　要：提出了一种基于类向量模型的中文姓名识别方法。该方法通过类向量的生成来模拟人工识别姓名的过程,采用V iterbi算法对未经切分的汉字串进行类向量标注得到类向量序列,通过检查相邻类向量中类别和向量分量的变化来最终识别出人名。该方法是完全数据驱动的,不需要姓名识别的模式和规则。通过对互联网上随机抽取的1 000篇文章进行测试,结果表明,中文姓名识别召回率为82.2%,准确率为70.3%。This paper presented an approach for Chinese names identification based on class vector model. The formation of class vector imitated human name identification. Character sequence was tagged using Viterbi algorithm with different class vectors. By checking the class of neighboring class vectors and the change of their components, the possible names were recognized without any template or rule. The test was carried out on real corpus from Internet. The experiments show that the precision and recall rate respectively reach 70.3% and 82.2%.

关 键 词：中文姓名识别 类向量模型 VITERBI算法 基于汉字 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]