The Trouble with What: Issues in method-independent task specifications
In this paper we discuss some issues concerning the organization of knowledge for reuse and we critically examine the ideas of knowledge separation and minimal ontological commitments.. Because knowledge structures can play multiple roles in a domain, it is not necessarily the case that search-control knowledge can be neatly separated from a domain ontology. This is particularly the case when only procedural descriptions of a task are available. Because expert knowledge is often `messy', clean separation of knowledge can be obtained only by removing the `troublesome' structures. However, we believe that this approach does not pay off in terms of reusability. Less knowledge means less reusability. We argue that a different approach to developing reusable ontologies is needed, which does not impose `strict' separation of knowledge, emphasizes the underlying assumptions about the available domain expertise, and does not trade knowledge for `formal purity'. These ideas are discussed in the context of the VT elevator design problem.
Accepted for publication at the ninth Knowledge Acquisition for Knowledge-Based Systems Workshop, Banff Canada, February 26-March 3, 1995.