The topic map model is a popular semantic web standard designed to organize contents from various resources and to enhance navigation in these resources. Building manually a topic map from a large content is impossible: a method supported by a tool is required. However, existing methods don't allow a construction of a contextualized topic map. This paper tackles this problem often encountered when organizing semantically heterogeneous contents. It proposes a model driven approach that builds a topic map to organize a semantically heterogeneous content. This approach uses UML profiles to describe the annotated content as well as the resulting Topic Map. It also uses the reification concept defined in the ODM (Ontology Definition Model) to reflect the semantic heterogeneity of resources to organize.