Developing internet usage and services urged play strong role for social network. Social networks are environment which uses internet as interface to provide relations between people, in the other word to interchange data and information between persons. Email and Instant Messengers are popular examples of them. Whereas these environments are continuously and instantly developing, revising and viewing by humans, they are good places for mining. In this paper, the topic of exchanged information between users in this type of networks will be our target. Our method is to use a hierarchical dictionary of semantically related topics and words that is mapped to a graph. Then extracted keywords from context of social network area compared to graph nodes and the dependency between them will be computed. This model can be used in many applications such as marketing, advertising and high-risk group detection.