基于混合CRFs&SVM应用于文本信息分类

Categorization of Text Information Based on CRFs&SVM

为了对文本信息进行更加有效分类,研究文本信息的一种新型分类。在了解文本信息分类和讨论条件随机场(CRFs)和支持向量机(SVM)的基础上,结合CRFs对上下文依赖性的优点以及SVM对高维问题的特殊解决方法,分析了二者结合的方法,并将其中一种结合方法应用到文本信息分类中。通过实验对此方法进行了验证。实验结果表明:CRFs&SVM方法的查全率、查准率以及F1-测量的各平均值比SVM方法都有所提高,具有良好分类性能和更加准确的分类率。进一步证明CRFs&SVM方法适用于文本信息分类,具有良好的应用前景。

In order to categorize text information more effectively,this paper adopted a novel method for categorization of text information by combining Conditional Random Fields(CRFs)of context dependence with Support Vector Machines(SVM)of the special solutions for high-dimensional problems.The results show that the average values of the recall,precision and F1-measure of CRFs&SVM method are higher than that of SVM method,presenting much more higher categorization accuracy and better categorization performance.In summary,the CRFs&SVM method is proper for the categorization of text information with wide application.