Website topics, describing aggregation and classification of website information, embodying information logic structure, is crucial for website information analysis. Analysis of logical structure is the reverse process of website design. In order to accurately analyze the site topics, the research proposed a topic structure model describing the organizational forms, logic relations and related properties of different website’s topics in a formal way, providing the necessary theoretical basis for the topic oriented web information extraction. On this basis, navigation-based topic structure induction was proposed with algorithm and experimental analysis to automatically construct topic structure of websites. Experimental results show that topic structure model generalizes most of the site’s topic structural characteristics, while the navigation based topic structure induction can correctly establish the site's topic structure, and has a faster running time.