Modeling Expert Control Knowledge.

Abstract : This research explored control knowledge in problem-solving in the domain of protein structure analysis. The objectives were to capture strategic knowledge for this problem. Implement this knowledge in working expert systems, and measure its effectiveness. We designed, implemented, and performed experiments on several control strategies for two aspects of the domain problem. The strategic knowledge was obtained by:1) active participation of chemists experts in the domain, 2) and by inferring strategic approaches from published papers and books. We tested our captured control knowledge by measuring its performance under several variations of the strategies to compare the efficiency of problem solving with different amounts and kinds of knowledge. We also compared results from our program with published results using other means. Important aspects of this work have been generalized to other problem domains. We find that much of the control knowledge developed from work on protein structures is easily applicable to other constraint satisfaction problems. Keywords: Control reasoning; Artificial intelligence; Blackboard architecture; Knowledge-based systems; PROTEAN.