Linguistic architecture

The highest level of articulation of the GENELEX model is between three layers:

• morphology
• syntax
• semantics

There is no entity like a ``lexical unit" as a whole in GENELEX: any lexical item can be seen either as a progression through the three layers or as a set of information regarding one given layer. Apart from tradition and intuition, the reasons for such a choice lay in what could be called the life cycle of a computational lexicon. A survey of lexical assets of GENELEX partners showed that every other database had reached a different step of development, and that stress had often been put on one of these layers, according to the needs for particular applications and/or to the human resources available. The organization was therefore defined so that it could allow the construction of a GENELEX lexicon to be undertaken starting from any layer (i.e. so that it could welcome data from the partners, after they have been checked with respect to the model of the relevant layer). This, even if the path from morphology to semantics through syntax may be seen as the most likely one. As a consequence, for example, no distinction related to polysemy is formally needed in a GENELEX lexicon untill the semantic layer is being constructed.

A unit of the morphological layer may have zero, one or more related units at the syntactic level ; a unit of the syntactic layer may have zero, one or more related units at the semantic level. A syntactic unit must be related to exactly one morphological unit, except for compound syntactic units, which are not directly related to the morphological units. A semantical unit may be related to one or more syntactic units, (when there are more, all these syntactical units must be themselves related to the same morphological unit). These constraints allow nevertheless a lexical database to focus on the semantic layer and to be more straightforward about the other ones, still being fully compatible with the GENELEX specifications.

In the next sections, a very short description of the type of information that may be found in the three layers is given. A graphical representation in the entity-relationship notation of GENELEX model is proposed in Annex number 5, under a rather general and simplified point of view.