International audienceIconic languages can represent concepts by the combination of graphical components (such as colors or pictograms). There are numerous examples, from traffic signs to computer user interface icons. However, these languages do not associate formalized semantics to their icons, which raises various problems: inconsistent combinations of graphical components, different interpretations of a given icon by two persons, difficulties to map the icons to the concepts of existing termino-ontological resources, etc. In this article, we describe a method that formalizes the semantics of an iconic language with an ontology. This method was initially developed for the VCM iconic language (Visualization of Concepts in Medicine), which enables to represent the main medical concepts (antecedents, disorders, treatments, etc.) by icons. We show that it can be generalized to other iconic languages, including traffic signs. We also describe four practical applications made possible by the formalization of the language semantics: the verification of icons consistency, the semi-automatic alignment with terminologies, the automatic generation of a pictogram lexicon and the automatic generation of icon labels. The article also presents the VCM ontology, the implementation details of a semantic iconic server with fast response times, and the evaluation results obtained when evaluating the four applications