摘要
文本分类存在维数灾难、数据集噪声及特征词对分类贡献不同等问题,影响文本分类精度。为提高文本分类精度,在数据处理方面提出一种新方法。该方法首先对数据集进行去噪处理,结合特征提取算法和语义分析方法对数据实现降维,再利用词语语义相关度对文本特征向量中每个特征词赋予不同权重;并利用经过上述处理的文本数据学习分类器。实验结果表明,该文本处理方法能够有效提高文本分类精度。
Text categorization faces the problems of dimensionality curse,noise data and different classification contributions for different feature words.In order to improve text classification accuracy,this paper presented a new approach to data processing.The approach first removed the noise data,and then employed feature extraction algorithms and semantic analysis methods to implement dimensionality reduction.Different weights were assigned to different text features based on a semantic similarity evaluation.The processed data were used to construct classifiers.Experimental results show that the text processing method can effectively improve the accuracy of text classification.
出处
《计算机应用研究》
CSCD
北大核心
2012年第12期4476-4478,共3页
Application Research of Computers
基金
国家自然科学基金资助项目(61170145)
国家教育部高等学校博士点专项基金资助项目(20113704110001)
山东省自然科学基金资助项目和科技攻关计划项目(ZR2010FM021
2008B0026
2010G0020115)
关键词
语义分析
降维
语义相关度
分类
semantic analysis
dimensionality reduction
semantic correlation
classification