Output multi-modal human-machine interfaces combine semantically output medias and modalities to increase interaction machine capabilities. In order to provide a rigorous development approach for these interfaces, we have proposed a generic formal model that formally describes the output multi-modal interface construction starting from the information generated by the functional core. This formal model is composed of two sub-models: the first one dedicated to the semantic fission of information and the second one relative to the allocation of modalities and medias for information. This paper presents an Event-B implementation of the semantic fission sub-model.