Many domains are facing challenging issues due to the volume, complexity and diversity of the digital information that they hold. Classical data integration techniques can hardly be appliedin this context to combine information residing at different sources and to provide the user with a unified view of the informationas they cannot cope with the complexity and the volume of the data. Among the proposals that were made in this domain, the dataspace paradigmiscurrently taking the first place in research. In this paper, we introduce a dataspace representation model. The model defines the dataspace as a set of classes, objects and relations. It addresses a number of shortcomings of existing models. In particular, it covers a large set of relation types, enabling the integration of many types of information such as structured and semi-structured data sources and ontologies. It also addresses the needs of multimedia data integration by providing means to represent similarity-based relations between multimedia objects.