Similarities

1. Levels
   The main entities of both models (i.e. GPMU, LU, UM, USYN, USEM) are seen as nodes of information related to content-wise descriptions or attributes that may therefore be widely shared by many items.
2. Morphology
   Both MULTILEX and GENELEX offer means to incorporate an explicit description of the morphological mechanisms that correspond to inflectional models (through the attributes and relations of MFG and MFP in GENELEX, through sets of declarations in the Rule Specification section of the dictionary in MULTILEX).
3. Theory-independence
   Both MULTILEX and GENELEX can welcome, within the same dictionary, different theoretical approaches and concurrent analyses. The representation language and the descriptive devices do not force the lexicographers to use one approach or another, provided that the pieces of information are translated into the specification. (However, the lexicographer has to control the consistency of his own choices.)
4. Morphology
   (GENELEX) and GPMU (MULTILEX) cover the same kind of information, i.e., morphology, orthography, phonology;
5. Compounds
   (GENELEX) and CGPMU (MULTILEX) have the same function;
6. Pragmatics
   In both models, 'pragmatic' information (in the sense of usage indication) can be attached to morphology, syntax and semantics, according to the needs
7. Syntax
   In both models, USEM and LU may be related to several syntactic descriptions.

As a provisory conclusion on this quick comparison, it can be stated that:

• On general methodological grounds, as well as in terms of relative role of the architectural proposal, the two projects may be viewed as quite similar.
• Node-wise organization of the lexical descriptions are based in principle on the traditional distinctions of descriptive levels, as they have been established in lexicography.
• There is no separate level of ``pragmatics", but pragmatic descriptions are ``distributed" over the partial descriptions from other levels.

However, the ``packaging" of information is different in the two proposals, which implies that relations between levels partly differ in the two models.
The next sections will discuss these approaches from the point of view of descriptive economy, power of generalization, strategies to accomodate even ``hard cases''. For the Architecture Task Group, a relevant question to answer will clearly be the status of syntactic information with respect to morphology on the one hand and to semantics on the other hand.