This paper presents a new approach to semantic recommender systems. The data analysed consists on a set of alternatives described by means of a list of tags. Each tag corresponds to one concept in a domain ontology. The first contribution of this work is the use of a multi-criteria outranking method to evaluate, rank and filter a set of alternatives. The second contribution is a procedure for exploiting the topology of the ontology to create a set of numerical criteria and represent the user's preferences about the concepts related to these criteria. The solution proposed reduces the computing time and converts a semantic problem into a multi-criteria numerical problem that can be solved using the ELECTRE method. The paper shows a case study in the field of touristic destinations recommendation with some tests with different user's profiles.