基于LDA和深度学习的文本分类方法  

Text classification method based on LDA and deep learning

作　　者：郑飞 韦德壕 黄胜[1,2] ZHENG Fei;WEI De-hao;HUANG Sheng(College of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Key Laboratory of Optical Communication and Networks,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065 [2]重庆邮电大学,光通信与网络重点实验室,重庆400065

出　　处：《计算机工程与设计》2020年第8期2184-2189,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61371096)。

摘　　要：针对文档集里的文本长度长短不一和特征提取困难等问题,提出一种基于LDA和深度学习的文本分类方法。结合LDA主题模型和Word2Vec词向量模型完成对文本词向量矩阵的构建,由结合融合层的卷积神经网络对构建好的词向量矩阵获取联合特征,将获取的特征送到softmax分类器得到分类结果。该方法在文本情感分类上进行实验,实验结果表明,该方法解决了文档集里的文本长度长短不一和特征提取困难等问题,在模型评价指标上都得到了提高。To solve the problems of different lengths of texts and difficulty in feature extraction in document collection,a text classification method based on LDA and deep learning was proposed.The LDA theme model and the Word2Vec word vector model were used to construct the text word vector matrix.The convolutional neural network combined with the fusion layer was used to obtain the joint features of the constructed word vector matrix,and the acquired features were sent to the softmax classifier to get the classification results.Experiments were carried out on text sentiment classification.Experimental results show that the proposed method solves the problems of different lengths of text and feature extraction in the document set,and increases model evaluation indicators.

关 键 词：文本长度 深度学习 词向量矩阵 卷积神经网络 情感分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]