<p>With the rapid development of social media, ethnic and religious public opinion—originating from geographically rooted issues—has increasingly spread and evolved online, resulting in more complex patterns of spatiotemporal evolution. Current research mainly focuses on dissemination and evolution mechanisms of public opinion information, with limited regard for the interactive evolutionary relationship and thematic features between offline geographical space and online network space. This study presents a framework for spatiotemporal evolution analysis and a latent topic analysis algorithm for public opinion on ethnic and religious affairs networks, considering both text space and geographical space, taking the example of ~25,000 pieces of all-media public opinion data on the “Ding Zhen popularity” event, using techniques such as spatial analysis, topic mining, and similarity comparison, explore the spatiotemporal distribution characteristics of online public opinion attention, analyze the content and intensity. The primary research findings indicate: 1. In terms of spatiotemporal evolution, network attention shows a prominent negative correlation pattern in space, hotspots are mainly concentrated in the central (e.g., Chengdu) and eastern (e.g., Beijing, Shanghai, and Guangzhou) regions, with network connection presents a “point axis” pattern; 2. Concerning topic distribution, the latent topic analysis algorithm proposed in conjunction with geographical locations effectively discovers the distribution of public opinion topics, topic intensity, and content evolution patterns. The average accuracy improved by 13% compared to the LDA algorithm; 3. Examining the latent topic features mapped from text space to geographical space, the similarity of public opinion topics is unevenly distributed at the urban scale, indicating that factors such as regional race, religion, and economic status have a significant impact on the distribution of ethnic and religious public opinion.</p>