基于多通道LSTM-CNN模型的Twitter情感分析  

Twitter Emotional Analysis Based on Multi-channel LSTM-CNN Model

作　　者：吉祥飞 李明东[1] 陶卫国 陈丽 JI Xiang-fei;LI Ming-dong;TAO Wei-guo;CHEN Li(China West Normal University,Nanchong 637000,China;Sichuan Delborui Technology co.,LTD,Nanchong 637000,China;Nanchong Civil Affairs Bureau,Nanchong 637000,China)

机构地区：[1]西华师范大学,四川南充637000 [2]四川德尔博睿科技股份有限公司,四川南充637000 [3]南充市民政局,四川南充637000

出　　处：《廊坊师范学院学报（自然科学版）》2019年第2期21-24,37,共5页Journal of Langfang Normal University(Natural Science Edition)

基　　金：四川省科技发展重点示范项目(川科高[2018]12,No.152)

摘　　要：Twitter作为人们表达观点的重要平台,已经成为文本情感分析的一个研究热点。文章提出一个基于多通道的LSTM-CNN模型,以twitter的tweets为研究对象,分析网络用户对该tweet的情感趋向,以验证该模型的有效性。引入深度学习理论,使用基于多通道的LSTM-CNN模型进行情感分析,以便更好地把握文本中的情感信息,提升情感分类的准确率。基于多通道的LSTM-CNN模型融合了CNN和LSTM的优点,在分析过长文本的情感特征时更加准确,适合twitter这种成段落的文本分析。实验结果表明,该模型在数据集上的准确率均优于SVM、LSTM和CNN。As an important platform for people to express their opinions,Twitter has become a research hotspot in text emotion analysis.This paper proposes a multi-channel lstm-cnn model,which takes tweets of twitter as the research object and analyzes the emotional trend of network users to the tweet to verify the validity of the model.The deep learning theory is introduced and the multi-channel lstm-cnn model is used for emotion analysis to better grasp the emotional information in the text and improve the accuracy of emotion classification.The multi-channel lstm-cnn model integrates the advantages of CNN and LSTM and is more accurate in analyzing the emotional characteristics of long texts,which is suitable for the text analysis in paragraphs like twitter.Experimental results show that the accuracy of the model is superior to SVM,LSTM and CNN.

关 键 词：情感分类 深度学习 TWITTER 多通道LSTM-CNN 情感特征 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]