ABSTRACT OF ENTERPRISE INTEGRATION USING THE 
SEMANTIC UNIFICATION META-MODEL (SUMM)

Jim Fulton

This paper describes an approach to enterprise integration 
using the Semantic Unification Meta-Model (SUMM), which is currently 
being developed by the Dictionary Methodology Committee (DMC) 
of the IGES/PDES Organization as a means of supporting the unification 
and integration of models based on the Standard for the Exchange 
of Product Model Data (STEP).

This approach assumes that unification must take place in a 
heterogeneous modeling environment in which there can be little 
if any change in the tools and languages being used by modelers.  
It also assumes that standardization should provide a pre-competitive 
framework within which vendors should be allowed maximum freedom 
to market solutions to communications problems.  Further, unification 
and integration techniques must be supported by technology that 
operates a PDES Level 3 (shared database) or 4 (shared 
knowledgebase); current Level 1 (file exchange) is acceptable only 
as a brief interim technology.  

The DMC believes that the most difficult problem involved in 
establishing the kind of communication needed to support enterprise 
integration is the semantic problem:  determining when the 
communicating parties are talking about the same thing, when 
they are not, and when they are talking about almost the same 
thing.  The determination of the precise degree of overlap in 
meaning of the expressions used by different groups is difficult 
even when they purport to use the same language.  When they use 
different languages, the problem has been almost insuperable, 
because there has been no standard meta-language, i.e., no language 
in which they could discuss precisely and rigorously the meanings 
of the expressions they use in their respective languages.

The SUMM is an attempt to address this lack of a meta-language.  
It is based on the classical predicate calculus.  Unlike other 
uses of the predicate calculus, the SUMM is not intended as a 
primary language for modeling domain knowledge.  Rather it is 
intended to provide a technique for rigorously defining the truth 
conditions of any model stated in any language that meets certain 
non-exclusive conditions:  that the information content of the 
modeling language is based on (1) the meaning of a set of 
"predicates" which classify objects and relationships, and 
optionally (2) the identification of individuals through the use 
of "terms".  The SUMM provides a meta-model for such languages that 
allows the expressions of each language to be mapped to a common 
ontology.  Such a meta-model can be developed at the instance level 
to support multilingual communication about particular individuals, 
at the model level to support multilingual communication about types 
of individuals, or, at the meta-model level to support multilingual 
communication about types of models.

This approach does not seem to require a standard concrete modeling 
language.  Any language that provides the abstract facilities defined 
for the SUMM can be used as a meta-language.  Such a language does 
have to be at least as rich and expressive as the languages being 
unified, and it must provide the facilities of the predicate calculus 
(truth functional combination and quantification) plus the ability to 
describe and identify sets (without yielding to the paradoxes of 
self-reference), models, states and events, as well as linguistic 
expressions themselves.  (Note that although having the apparatus 
of the predicate calculus is a necessary condition, it is not 
sufficient:  The language must also have the predicates appropriate 
to the description of the syntax and semantics of linguistic 
expressions.)  

Each of these abstract requirements can be implemented in a wide 
variety of concrete languages.  This means that a unification 
language can be suited to the backgrounds of the enterprises 
that have identified a need to communicate in spite of linguistic 
barriers.  Presumably there needs to be at least one recognized 
concrete implementation of the SUMM, and probably more than that.  
Although there does not yet seem to be a language that fully 
addresses all the requirements identified in the SUMM, it appears 
that there are several existing languages that might be extended 
to meet the unification needs of particular communities:  a SUMM 
version Express might address the needs of the STEP community, a 
SUMM dialect of SQL might support the relational database community, 
and a version of KIF with the appropriate ontology might support 
the knowledge base community.  Once a basic concurrence on this 
approach is achieved, an exploration of these concrete 
implementations might profitably proceed in parallel with the 
refinement of the abstract definition of the SUMM.