In this paper, we introduce a Media Decision Taking Engine (MDTE), enabling the automatic selection and/or rating of multimedia content versions, based on the available context information. The presented approach is fully semantic-driven, which means that we not only semantically model the context information, but also the decision algorithms themselve, which are represented in N3Logic, a rule language that extends RDF. The decision rules are based on a rating function, supporting the specification of weights and affinity parameters for each environment property. Finally, we show how the MDTE is integrated in a media delivery platform, using the provisions of the existing Web infrastructure.