page 4  (14 pages)
to previous section3
5to next section

system in terms of its rational behaviour: why does the system (the agent") perform this action", independent of its symbolic representation in rules, frames or logic (the symbol level").

2.2 Multiple views on medical domain knowledge During the last decade, a number of proposals have been put forward in knowledge engineering research for describing knowledge-level models [Clancey, 1985; Neches et al., 1985; Wielinga & Breuker, 1986; Chandrasekaran, 1988; Marcus, 1988; Musen, 1989; Steels, 1990; Schreiber et al., 1993; Wielinga et al., 1993; Ramoni et al., 1992]. A common distinction that is being made is between (i) domain knowledge, defining a declarative theory of the application domain, and (ii) control knowledge which specifies how to use domain knowledge to solve a problem.

In this paper we focus on the description of domain knowledge and its relation with control knowledge. Although terminology varies, there appear to be at least three descriptive levels of domain knowledge that can be found in most approaches:

Application ontology
The application ontology, in short ontology" 1, specifies the structure of the domain knowledge in terms of a number of (domain) knowledge types. It characterises the types of objects and expressions that one finds in the domain knowledge.

manifestation of diseasefinding

evoking
strength frequency
importance
function function function

class relation class

FIGURE 1: Example domain knowledge types, derived from the ontology underlying internist-1 [Miller et al., 1982]

Fig. 1 shows a partial example ontology, derived from the way in which domain knowledge is described in internist-1 [Miller et al., 1982]. As [Davis et al., 1993] point out, there is an additional underlying ontological commitment in such an ontology, namely the representational primitives that are used for describing the ontology. In games we have chosen the primitives used by Gruber in Ontolingua [Gruber, 1992]: classes, relations and functions.

Application knowledge
The application knowledge constitutes the actual" domain knowledge. Roughly, it consists of two parts:

1In a previous publication [Wielinga et al., 1992] the first author favoured the term schema" above ontology" to stress that it is product of engineering, and does not necessarily imply a faithful description of the real world. The use of the second term has, however, become predominant in the literature.