锐化高斯模板在文本特征项权重调整方法中的应用

An Application of Sharpen Gaussian Template in a Text Feature Weight Adjustment Methodology

阐述二维高斯模板和锐化高斯模板的构建过程及其对图像的锐化处理技术,提出文本特征项权重调整方法的主要思想,并以此为依据,论述基于锐化高斯模板的文本特征项权重调整方法。以搜狗自然语言实验室的文本分类语料库中的新闻文档作为实验对象,采用宏平均F测度值作为分类效果评价标准,采用中心点法和K-最近邻方法(KNN)作为分类器的分类算法进行实验。实验结果表明:在KNN分类法下,基于锐化高斯模板的文本向量模型权重调整方法起到提升分类效果的作用,但对中心点法的影响并不大。

This paper introduces Gaussian Template and Sharpen Gaussian Template in computer image processing tech- nology and summarizes main ideas of text feature weight adjustment, then proposes a text feature weight adjustment method- ology based on Sharpen Gaussian Template. With corpus of Sogou Lab Data, KNN classifier and Class - center classifier, this methodology is experimented by Macro - averaging F - measures. The experimental result shows that the KNN classifi- er with this methodology performs better than the traditional method. However, Class - center classifier with this methodol- ogy has no significant improvement.