Our first level of representation is called RROM (Relational Ressource Organisation Model). A RROM is composed of a set of Ressource Entities (RE) and a set of relations between these entities. A RE corresponds to an independent and abstract type of data that is used in a NLP system (for example word, lemma or category). Given a set of ressources, Independant data means that this data is not the result of a set of relations between other RE. A RE is represented with a general name and is associated to a data type definition. An instanciation of a RE is a realization of this RE according to the corresponding data type specifications. In following figures, an RE is graphicaly represented with a square box.
The relations between entities used in this model are characterized by two couples of integers on each edge. Depending on the direction of the relation, this couple gives the arity of the relation with the RE given by the edge, by analogy to the couples on the edges used in relational databases entity/relation models. Two RE can also be in relation. A RROM can be graphically represented with diagrams describing which REs are related to one another. In these diagrams, a Ressource Relation (RR) is represented with ellipsis. We distinguish two kinds of edges: Unary edges (single line) which indicate a single link relation and n-edge (double line) which means that a relation can link n instatiations of a RE at the same time.