三种方法的NP识别比较实验研究

Comparative Experimental Research on Three Methods of NP Recognition

利用错误驱动法、支持向量机法和隐马尔可模型三种方法对汉语文本进行名词短语识别,对实验进行比较分析,结果表明SVM与HMM的识别效果总体上要好于错误驱动法,HMM法在封闭测试中优势明显.研究表明错误驱动法应用于解决从语料库中学习转换规则的传统问题;SVM方法适用于解决两类别的分类问题;而HMM方法侧重应用在与线性序列相关的现象上.

Using three methods of error- driven, support vector machine and hidden markov model, noun phrase recognition is carried on to chinese text, through comparative analysis to experiment, the results indicate that the recognition effects of SVM and HMM are overall better than the method of error- driven, HMM method has the distinct advantage in the dosed test. The research indicates that SVM method is suitable to solve classification problems of two categories, but HMM method is suitable to apply on the correlative phenomenon of linear serial.