基于词性过滤和改进边权重的短文本分类算法  

A short text classification algorithm based on part of speech filtering and improved edge weight

作　　者：许梦玥 侯秀萍[1] 王俊华[1] XU Mengyue;HOU Xiuping;WANG Junhua(School of Computer Science&Engineering,Changchun University of Technology,Changchun 130102,China)

机构地区：[1]长春工业大学计算机科学与工程学院,吉林长春130102

出　　处：《长春工业大学学报》2023年第6期546-551,共6页Journal of Changchun University of Technology

基　　金：吉林省教育厅“十三五”科学技术项目(JJKH20191311KJ)。

摘　　要：针对短文本存在特征稀疏和信息不规范等特点,文中在TextGCN模型的基础上通过增加词性过滤减弱无关词对特征选择的影响,并加入TF-CR算法提高类别无关词权重,最后,通过与几个经典模型进行对比,验证改进模型的有效性。In view of the features of the short text such as sparse features and non-standard information,this paper,based on the TextGCN model,reduces the influence of irrelevant words on feature selection by adding part of speech filtering,and adds TF-CR algorithm improves the weight of category independent words,and finally verifies the effectiveness of the improved model by comparing with several classical models.

关 键 词：词性过滤 特征选择 短文本分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]