Abstract: To improve interaction efficiency and reduce interactive burden for browsing massive social information on the Internet, a multi-cue information organizing method and a LDA-based topic extraction approach are proposed. A large amount of messages of a location are aggregated to discover the topics, and then the topic of each message is selected using generative probabilistic model. A visualization method is designed and implemented to display social information in virtual scenes based on multi-cue information organizing, which utilizes classical visual structures and natural interaction. Experiment results prove the effectiveness of the topic extracting method. Interaction efficiency is improved effectively and different requirements for browsing information are met with the multi-cue information organizing and visualization method.