摘要
为了更好地获悉社区居民在网络中所反映的民生问题,把握亟需关注的热点,基于便捷高效、沟通互动性强的在线聊天数据,提出了一种基于情感与热点话题的综合分析模型。首先,采用半监督的情感标注模型与基于注意力机制的双向长短期记忆网络模型对社区相关数据进行居民情感分析;其次,通过隐狄利克雷分布主题模型对热点问题进行研究;最后,结合话题类别与情感分布进行综合分析。实验结果表明,采用半监督的情感分类模型最终分类准确率可达到89.92%,相较于其他基线模型,取得了更好的分类效果。经卡方检验后可知热点话题与情感分布之间具有相关性,不同社区的居民关注的话题、发言的数量及发言的长度等均存在较大的差异,各社区集中讨论的时间点与其从事职业具有密切关系。这些均可为居民社区服务部门、社区治理部门及相关社会工作者的工作提供切实有效的参考依据。
To better understand the livelihood issues reflected by community residents on the Internet and grasp the hot topics that need urgent attention,a comprehensive analysis model based on sentiment and hot topics is proposed by using convenient,efficient,and communicable interactive online chat data.Firstly,a semi-supervised sentiment annotation model and an attention-based bidirectional long-short term memory network model are used to analyze community-related data for resident sentiment,followed by a latent Dirichlet allocation topic model for hot issues,and finally,combining topic categories and sentiment distribution be explored.The experimental results show that the final classification accuracy of the semi-supervised sentiment classification model can reach 89.92%,which achieves better classification results than other baseline models.A chi-square test shows that there is a correlation between hot topics and the distribution of sentiment.There are significant differences in the topics of interest,the number of statements and the length of statements made by residents in different communities,and the point in time when discussions are concentrated in each community is closely related to their occupations,which can provide a valuable reference for the work of community service departments,community governance departments,and relevant social workers.
作者
蔡云戈
范永胜
冯骥
CAI Yun-ge;FAN Yong-sheng;FENG Ji(School of Computer and Information Science,Chongqing Normal University,Chongqing 401331,China)
出处
《计算机技术与发展》
2023年第5期42-48,共7页
Computer Technology and Development
基金
重庆师范大学(人才引进/博士启动)基金项目(17XCB008)
教育部人文社会科学研究项目(18XJC880002)。
关键词
居民社区
情感分析
在线聊天
热点话题
注意力机制
residential communities
sentiment analysis
online chat
hot topics
attention mechanism
