Identification of communities in social networks has become a hot research topic in recent years. Many algorithms have been designed to discover networks’ community structure. Most of these algorithms detect disjoint communities, which means that every community member belongs to a single community. These models do not consider that a person may have more than one interest. Thus, lately, a few methods have been designed to find overlapping communities. But most researchers have either emphasize to solve this problem on computing network’s structural properties, or using graphical models for the community extraction process, where structural properties of networks are not considered. However, when end users are connected with each other by documents, posts or comments it is not possible to ignore underlying informations’ semantics from these texts. In this paper, we propose a novel approach to combine traditional network analysis methods for overlapping community detection with topic-model based text mining techniques.