【作者】 林游龙； 余智华； 程学旗； 刘悦；

【Author】 Youlong Lin~(1,2),Zhihua Yu~1,Xueqi Cheng~1,Yue Liu~1 1 Institute of Computing Technology,Chinese Academy of Sciences,Beijing,10080 2 Graduate University of Chinese Academy of Sciences,Beijing,100190

【机构】 中国科学院计算技术研究所； 中国科学院研究生院；

【摘要】 基于向量空间模型的分类方法是目前各种分类方法广泛使用的文档结构表示方法,在对基于向量空间模型的分类方法的研究发现,基于向量空间模型的分类方法存在不合理之处,即特征值之间的"鸿沟",这种鸿沟会导致向量空间模型中两点之间的距离的计算出现偏差,本文介绍了一种使用虚点的方法,这种方法消除了特征值之间的鸿沟,使得分类的效果得到了提高。该方法是通过重新定义特征权重,调整向量空间模型中点的特征值,即相当于重新定义向量空间中的点,这样的点是相对于原来向量空间模型中的点的矫正映射,即就好像是虚拟点一样,最后问题归结为计算向量空间模型中的点与虚拟点的映射函数。理论分析表明虚点方法能提高基于向量空间模型的分类方法的效果,在SVM中运用虚点方法的实验结果表明,运用虚点方法的SVM的精确度得到了提高,这种结果验证了本文提出的虚点方法的有效性。更多还原

【Abstract】 Vector space model(VSM) is the widely used model in the representation of the document structure in a variety of classification methods.The research on the vector space model based classification method shows that there is unreasonable point,that is,the gap between the features,this gap will lead to the deviation in the calculation of the distance between the two points in vector space model.This paper proposes a method of the virtual point to eliminate the gap between two features which improve the performance of the text categorization. The method is to adjust the feature value of point in the vector space model by redefining the weight of feature which is equivalent to the redefinition of the point in the vector space.Compared with the point in the original vector space model,the point is assumed to be the correctly mapping,that is,like a virtual point.Finally the problem boils down to the calculating of the mapping function between the vector space model and of virtual vector space model.Theoretical analysis showed that the virtual-point method can improve the performance of text categorization based on the vector space model.The experimental results of the support vector machine categorization method using virtual-point show that the performance has been improved,which verify that the virtual-point method is effective.更多还原