摘要
随着移动互联网的快速普及,网络上各种电商平台的商品评论、社交网络平台的社交评论等激增,这些评论信息中的情感倾向具有巨大的商业价值和社会价值。利用图卷积网络(GCN)技术对微博评论数据集和酒店评论数据集等中文短文本进行情感分析,分析过程中对相关评论文本进行情感标注、文本清洗、中文分词等预处理工作,再使用word2vec模型对预处理后的评论文本进行文本向量化,并利用GCN模型与深度学习模型CNN、RNN、RCNN等进行情感分析对比实验。实验结果表明,本文提出的基于GCN的细粒度模型相比于CNN、RNN、RCNN等模型能提高评论类文本情感分析的准确率,在微博评论和酒店评论的数据集上的准确率分别提高了4.79%、1.58%以上,从而验证了该模型在中文短文本情感分析方面的有效性。
With the rapid popularization of mobile Internet,the number of product reviews on various e-commerce platforms and so-cial reviews on social network platforms is increasing rapidly.The emotional tendency in these reviews has great commercial value and social value.In this paper,graph convolution network(GCN)technology is used to analyze the sentiment of short Chinese texts such as microblog comment data set and hotel comment data set.In the process of analysis,sentiment annotation,text cleaning,Chi-nese word segmentation and other preprocessing work are carried out for relevant comment texts.Then word2vec model is used to quantify the preprocessed comment texts,GCN model and deep learning model CNN,RNN,RCNN are used to carry out the com-parative experiment of emotion analysis.Experimental results show that the proposed fine-grained model based on GCN can im-prove the accuracy of sentiment analysis of comment text compared with CNN,RNN,RCNN and other models,and the accuracy of sentiment analysis of microblog comments and Hotel Comments is improved by more than 4.79%and 1.58%respectively,which verifies the effectiveness of the model in sentiment analysis of Chinese short text.
作者
陈俊涛
刘力铭
车月琴
CHEN Jun-tao;LIU Li-ming;CHE Yue-qin(Guangzhou City Polytechnic,Guangzhou 510405,Guangdong)
出处
《电脑与电信》
2023年第3期79-84,共6页
Computer & Telecommunication
关键词
图卷积网络
词向量
情感分析
CNN
短文本
graph convolution network
word vector
sentiment analysis
CNN
short text
