A two nodes model

The MULTILEX linguistic architecture is based on the independant description of, on the one hand, the words as they behave in the oral and written codes and, on the other hand, the meanings expressed by the words within a language. Lexical information is divided in accordance with this distinction and distributed around two nodes.

1. One node - called GPMU (Graphic-Phonologic-Morphological Unit) - is a triple identifying one set of coherent orthographic, phonological and morphological behaviours belonging to a canonical form.

   The canonical form is used in MULTILEX as an organising principle as one may speak of the different meanings of a word because a meaning is always associated to a form. The orthographic standard form is fixed in our languages by convention, apart from a number of exceptions (for example, colineau or colinot in French, or yoghurt, yogurt or yoghourt in English). In these cases, each form is covered by a distinct GPMU.

2. The other node - called LU (Lexical Unit) - identifies a single meaning within a language.

Lexical description results from the linking of the appropriate GPMUs and LUs. (see example: Annex number 1 ``THE BALLE SCHEMA")

• several LUs can share the same GPMU (this case accounts for homonymy)
• one LU can be linked to several GPMUs (this case accounts for variants)
• special behaviours cause a GPMU to be created (this case accounts for restricted usages).