In this paper, I describe an approach for rule-based systems exploitin
g a purely object-oriented approach. The innovative idea is to conside
r a new kind of relation between knowledge objects which are believed
to model implication relations of production rules. A knowledge base c
onsisting of such objects can be conveniently represented as a Marker
Propagating Graph (MPG), which provides rule-based-like representation
features. The inference is seen as a marking propagation through the
graph. This approach preserves the best of both the object orientation
features and expert system functionality. This experimental study con
cerns the design and the implementation of a medical system for automa
tic interpretation of biological tests in Preventive Medicine Centers.
Because the use of such a system is planned for many years, the possi
bilities for its future extensions are seriously considered. This woul
d not be possible without a good appreciation of object orientation fe
atures. (C) 1998 Elsevier Science Ltd. All rights reserved.