This paper describes an approach for semantic description and retrieval of multimedia data described by means of MPEG-7. This standard uses XML schema to define the descriptions. Therefore, it lacks ability to represent the data semantics in a formal and concise way and it does not allow integration and use of domain specific knowledge. Moreover, inference mechanisms are not provided and hence the extraction of implicit information is not (always) possible. To address these issues, we propose to add a conceptual layer on top of MPEG-7 metadata layer, where the domain knowledge is represented using a formal language. A set of mapping rules is proposed.