摘要
2019年新型冠状病毒(COVID-19)肺炎疫情对人民生产生活各方面产生严重影响,为协助政府把握社会舆论,更加科学有效地做好预防控制工作的宣传和舆论引导,本文以与新型冠状病毒肺炎疫情相关的微博内容为研究对象,采用深度学习技术对网民情感进行分析和识别。首先,采用文本分词、正则表达式、词性和停用词表过滤等方法进行预处理操作;其次,构建Bert-CNN模型和Bert-RCNN模型对微博中的网民情感进行识别;最后,验证两种模型的识别效果,并与其他模型进行对比分析。实验结果表明,Bert-RCNN模型的效果最好,其F1-score值为0.702、准确率为73.56%。
The COVID-19 pandemic in 2019 has attracted widespread attention in China.In order to assist the government to grasp the real public opinion and conduct publicity and public opinion guidance for prevention and control work in a more scientific and effective way,based on the micro-blog content related to COVID-19,this paper adopted the deep learning technology to classify and predict the emotions of netizens.Firstly,text segmentation,regular expression,part of speech and stop vocabulary filtering methods were used for preprocessing.Secondly,Bert-CNN model and Bert-RCNN model were constructed to classify and predict the emotions of netizens in micro-blog.Finally,the identification effects of the two models were verified and compared with other models,and the experimental results showed that the Bert-RCNN model achieved the best effect,with F1-score of 0.702 and accuracy of 73.56%.
作者
张苑
祝小兰
杨东晓
ZHANG Yuan;ZHU Xiaolan;YANG Dongxiao(Department of Computer Technology and Applications,Qinghai University,Xining 810016,China)
出处
《智能计算机与应用》
2022年第3期40-45,52,共7页
Intelligent Computer and Applications
基金
国家自然科学基金(61866031)
青海省科技厅青年自然科学基金(2021-ZJ-952Q)
四川省科技厅重点研发项目子课题(2020YFS0575)
作者简介
张苑(1999-),女,本科生,主要研究方向:深度学习、自然语言处理;通讯作者:祝小兰(1990-),女,硕士,讲师,主要研究方向:机器学习、信息安全,Email:zxlanscu@126.com;杨东晓(1994-),男,硕士,助教,主要研究方向:时空序列数据挖掘。
