The use of ontology for knowledge organization is a common way to solve semantic heterogeneity. But when the categories of the relative concepts of ontologies are different, the semantic interoperation will encounter new obstacles. A new theory model is needed to solve this kind of problem. Information flow (IF) theory seems to be a promising avenue to this end. In this paper, we analyze semantic heterogeneities among two data sources in distributed Xu Beihong’s Galleries and describe an instance based approach to accomplish semantic interoperation and IF based formulation process of the global semantics. The emphasis is how to use IF theory to discover concept and instance duality in knowledge sharing. In the end, the more interrelated problems are farther discussed.