Abstract: In recent years, web data mining has became increasingly important in software-based teaching. This paper takes the Bilibili screen video network as the platform, takes the Python crawler as a tool to collect a large amount of barrage data, and conduct mining analysis to indirectly evaluate the video content. Firstly, the research shows that the most popular part of videos based on the time distribution of barrages appearing. Secondly, it analyzes the most interesting part of the video based on the time distribution of the barrage in the video. Then, combined with the sentiment orientation analysis tool in Baidu AI natural language processing and one-way analysis of variance in MATLAB, the characteristics of the barrage at the beginning of the video are studied. Thirdly, the AI language sentiment analysis and keyword analysis methods are used to study the emotions and categories of the barrage, and then the content characteristics of the video are analyzed. The final result of the experiment shows a unique evaluation of popular short videos, which has certain reference value for short video authors and platforms. The barrage-based research method also provides a new idea for the automatic identification and evaluation of video content.