Abstract: Empirical research on network public opinion analysis is mostly conducted on a single media, with high time cost and strong subjectivity. In the context of the rapid development of media integration in the post-epidemic era, cross-media and multi-terminal heterogeneous information brings new challenges to the analysis and governance of online public opinion. This research focuses on the public opinion problems in the process of epidemic control in colleges and universities, and proposes a public opinion sentiment analysis method based on word frequency analysis and LDA model. Collect public opinion data from August~October 2021 that are too “closed for school opening” and “National Day holiday as scheduled” through web crawlers. After data cleaning and word segmentation, the TF-IDF algorithm is used to extract high-frequency words and build a word cloud, through the LDA model to extract the topic of public opinion and then conduct public opinion analysis. Verified by cross-media public opinion examples of epidemic control in colleges and universities, the results show that this research can enrich the government response system for public crises, and provide theoretical basis and technical support for effective government management.