基于图小波网络模型的文本分类研究  

Research on text classification based on graph wavelet network model

作　　者：马诚[1] 贾凯莉 李云红[1] 高子明 候嘉乐 MA Cheng;JIA Kaili;LI Yunhong;GAO Ziming;HOU Jiale(School of Electronics and Information,Xi’an Polytechnic University,Xi’an 710048,China)

机构地区：[1]西安工程大学电子信息学院,陕西西安710048

出　　处：《电子设计工程》2022年第11期17-21,共5页Electronic Design Engineering

基　　金：西安市科技局高校人才服务企业项目(2019217114GXRC007CG008-GXYD7.2,2019217114GXRC007CG008-GXYD7.8)。

摘　　要：针对文本分类中获取文本复杂特征困难、分类准确率低等问题,建立基于图小波网络文本分类模型。根据语料词库共现信息及词与文档的关系构建文本图,使用改进TF-IDF算法、PMI算法计算词与文档之间和词与词之间文本图的权重;建立基于图小波文本分类模型,将构建的文本图输入到GWNN模型中。经R8、R52及Ohsumed英文语料库测试结果表明,文本分类准确率分别达到98.09%、93.91%及69.3%,验证了基于图小波网络模型的有效性,也为文本分类提供了一种有效的方法。Aiming at the problems of limited ability to characterize complex features of text and low classification accuracy in text classification,a text classification model based on graph wavelet network is established.The text graph is constructed based on the corpus word co-occurrence information and word-document relationships,and the weights of the word-to-document and word-to-word text graphs are established using the improved TF-IDF algorithm and PMI algorithm.A graph wavelet-based text classi-fication model is established,and the constructed text graph is input into the GWNN model.The test results of R8,R52,and Ohsumed English corpus showed that the text classification accuracy reached 98.09%,93.91%and 69.3%,respectively,which verified the effectiveness of the graph wavelet network-based model and also provided an effective method for text classification.

关 键 词：PMI算法 改进TF-IDF算法 图小波网络 文本分类 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]