【摘要】 本文以社会化问答社区为例，探究疫情期间健康信息需求的主题与用户情感变化特征，以期改进问答社区在突发事件中的应急策略，通过数据采集和清洗、文本预处理、LDA主题模型、BERT+BiLSTM情感分类模型对25 540条数据进行知识挖掘和主题-情感协同分析。研究结果显示，本文使用的方案能够有效捕捉疫情期间网民需求健康信息的主题特征。在情感分类方面，BERT+BiLSTM模型的分类准确率较基线模型提升了11.75%。为更好地应对突发公共卫生事件，本文建议社会化问答社区应自行生产科学的健康信息、提高针对主题的舆情监控力度并积极引导用户认知。更多还原

【Abstract】 Taking the socialized question and answer community as an example, this paper explores the theme of health information needs and the characteristics of user emotion changes during the epidemic, with a view to improving the emergency strategy of the question and answer community in emergencies. Through data collection and cleaning, text pre-processing, LDA topic model, BERT+BiLSTM sentiment classification model, knowledge mining and theme emotion collaborative analysis are conducted on 25 540 pieces of data. The research results show that the scheme used in this article can effectively capture the thematic characteristics of the health information needs of netizens during the epidemic period. In terms of emotion classification, the BERT+BiLSTM model has improved the classification accuracy by 11.75% compared to the baseline model. In order to better respond to sudden public health incidents, socialized Q&A communities should produce scientific health information on their own, improve public opinion monitoring targeting the theme, and actively guide user awareness.更多还原