CONCEPTUAL AND FORMAL SPECIFICATIONS OF PROBLEM-SOLVING METHODS

Reusable problem-solving methods as provided by the PROTEGE-II improve knowledge engineering by allowing developers to design reasoners quic kly from pre-existing components. The PROTEGE-II approach allows devel opers to select methods from a library, and to map the methods to a do main ontology. Still, these methods lack a clear conceptual and formal description that would enable their reuse through matching their comp etence and assumptions with the available domain knowledge and the giv en task. KARL is a conceptual and formal knowledge-specification langu age that provides modeling primitives for specifying problem-solving m ethods. In this paper, we show how the code and informal descriptions of problem-solving methods in PROTEGE-II can be complemented with the conceptual and formal method definitions in KARL. For our case study w e choose two methods from the PROTEGE-II framework: chronological back tracking and a task-specific refinement, the board-game method. In add ition to the conceptual and formal specification of these methods, we provide insights in the refinement of general-purpose methods to task- specific (i.e., strong) problem-solving methods. We further show how a task-specific method can be adapted to a given domain and application . In the case of both methods, we achieve this adaptation by introduci ng ontological commitments over the terminological structure of the en tities used to describe the states of the reasoning process, and by us ing these terminological structure to define state transitions of Larg er grainsize.