Abstract: 【Purpose/significance】Accurate and efficient online public opinion monitoring enables university managers to quickly and accurately grasp the trend of online public opinion, and provide support for their analysis, judgment, and formulation of countermea⁃ sures. It is very important to resolve the crisis of internet public opinion in universities, maintain the safety and stability of the campus, ensure the order of teaching and scientific research, and build a harmonious campus.【Method/process】This paper collected the theme posts and replies about Jilin University from Zhihu and Baidu Tieba in the autumn semester of 2021 as data sources. The model based on Albert and Bilstm-Attentionis constructed to identify positive, neutral, and negative emotions in the text, focus on extracting hot topic information with negative emotional tendency, analyze the reasons, and propose solutions.【Result/conclusion】Results show that the model can effectively identify the emotional tendency of popular topic information and track the trend of monitoring data in time. This provides an effective method and model basis for university administrators to monitor network public opinion. Compared with previous models, this model can more fully capture the semantic information contained in the public opinion text.【Innovation/ limitation】The themes of Internet public opinion in universities are diverse, and the text structure is complex and novel. Given these characteristics, the attention mechanism can deeply mine the paragraphs with emotional tendency features contained in the text, so that the model can more accurately predict the emotional tendency of online public opinion. However, the public opinion monitoring model proposed in this paper does not take into account the fluctuation of public opinion development under time evolution, and the model will be further optimized to achieve high-precision dynamic detection in future research.