With the emergence of Web 2.0, the volume of multimedia documents, particularly the socio-tagged images, has become very considerable. This has made it the annotation and interrogation process of this type of documents a consuming and elusive task. A promising solution is to design methods allowing to model the semantic content of socio-tagged images in order to facilitate their annotation and the research process. Thus, we present in this paper a conceptual modeling approach of the semantic content of these images. First, our idea consists in expanding the conceptual graphs formalism in order to represent the relationships between the concepts and those that are between the concepts and their properties. Second, we use the extended graph to model the semantic content of the socio-tagged images. Experimental studies are conducted on a collection of 25.000 socio-tagged images shared in Flickr. The results demonstrate the effectiveness of our proposed approach.