网络舆情监测综合实验设计

Comprehensive experimental design for network public opinion monitoring

网络自媒体海量舆论信息转发、情感两极分化、谣言混杂等极易引发网络舆情事件,对网络舆情监测综合能力的培养提出了更高的要求。针对现有教学中普遍缺乏综合实验这一实际情况,设计了一套网络舆情监测综合实验方案。首先,以成果导向教育理念为引导,确立了实验教学目标,规划了实验内容框架。然后,详细设计了从舆情文本获取到预处理,从主题挖掘到以基于变换器的双向编码器表征(bidirectional encoder representation from transformers, BERT)大模型为核心的情感分析、谣言检测等一系列实验模块。最后,使用网络舆情监测实例展示了实验效果。教学实践表明,实验方案能够使学生深入掌握网络舆情监测基本流程和实操技能。

[Objective]With the rapid development of various online We Media applications,monitoring and guiding internet public opinion information are not easy.The massive amount of forwarded public opinion information,emotional polarization,and mixed rumors in online We Media are prone to trigger online public opinion events,posing higher requirements for developing a comprehensive ability for online public opinion monitoring.In response to the lack of comprehensive experiments in the existing teaching methods,a comprehensive experiment scheme is designed for monitoring online public opinion.[Methods]First,guided by the concept of outcome-based education,the experimental teaching objectives are established,and the experimental content framework is planned with three major categories of experiments.Subsequently,by applying the information ecosystem and lifecycle theory,a spatiotemporal element identification experiment on online public opinion dissemination is conducted to address the issues of monitoring time and spatial scope and to minimize the impact of the massive and high-speed nature of public opinion data on monitoring methods.The collection and preprocessing experiments on public opinion text data are completed using web crawlers and text vectorization methods to obtain efficient vector representations of public opinion text.In this step,students can experiment with the customized crawler method to understand the crawler design steps and how to flexibly respond to anticrawling measures.Moreover,students are guided to complete experimental processes such as text data cleaning,word segmentation,part-of-speech tagging,stop word filtering,and vector representation.Further,text mining helps to quickly discover potential topic viewpoints from massive text data,therefore,it is an important approach for efficiently observing the evolution of public opinion on a particular topic.Two types of experiments are organized based on text clustering methods and latent Dirichlet allocation methods.Considering that sentiment analysis and rumor detection are important tasks in online public opinion monitoring,an experimental module is designed based on the BERT model to guide students to use domain-specific datasets to fine-tune the whole model after introducing three network layers,namely TextCNN,TextRCNN,and Bi LSTM+Attention,to train the transfer and application ability of BERT pretrained models in sentiment analysis and rumor detection tasks.[Results]Finally,online public opinion monitoring of Typhoon“Doksuri”is used as an example to illustrate the detailed process involved in this comprehensive experiment.The experimental results indicate that:①the development of Weibo public opinion on Typhoon“Doksuri”comprised the information lifecycle theory.②LDA theme analysis shows that the various stages of public opinion development on Weibo during Typhoon“Doksuri”conform to people’s desire for information during the typhoon period.③The evolution detection of emotional tendencies found that the development of public opinion and emotional tendencies in Typhoon“Doksuri”agrees with the overall trend of public opinion.The dominance of positive emotions indicates public confidence in overcoming the typhoon,while the dominance of negative emotions in the later stage of public opinion development reflects people’s concerns regarding disaster relief and secondary disasters.④The rumor detection results indicate that more blog posts express true information on“Doksuri,”indicating that people hope to convey more real information during the critical fighting period against typhoons.Finally,strategies for public opinion control were suggested to help students gain a deeper understanding of the significance of public opinion monitoring.[Conclusions]The comprehensive experiment on network public opinion monitoring designed in this article can meet the needs of relevant experimental teaching courses.The teaching practice shows that this experimental scheme can enable students to deeply grasp the basic process and practical skills of online public opinion monitoring.