Journal: International Journal of Expert Systems: Research & Applications
Reusable problem-solving methods improve knowledge engineering by allowing developers to design reasoners quickly from pre-existing components. An important factor for successful method reuse is the specification of preexisting methods and support in their task- and domain-specific refinements. The PROTÉGÉ-II approach allows developers to select methods from a library, and to map the methods to a domain ontology. KARL is a formal and executable knowledge-specification language that provides modeling primitives for expressing problem-solving methods. In this paper, we show how the code and informal descriptions of problem-solving methods in PROTÉGÉ-II can be supplemented with the formal method definitions in KARL. For our case study, we have chosen two methods from the PROTÉGÉ-II framework: chronological backtracking and its task specific refinement: the board-game method. In the paper, we give insights in the refinement of general-purpose methods to task-specific (i.e., strong) problem-solving methods. We show further, how a task-specific method can be adapted to a given domain and application. In both cases this is achieved by introducing ontological commitments over the terminological structure of the entities used to describe the states of the reasoning process of a method.