(A) study on knowledge refinement technique for an oriental medicine knowledge base

Abstract

New efficient approach was conceived and tested on the expert systems of oriental medicine. The knowledge bases of these systems were constructed with two knowledge sources; the rough knowledge was obtained from a patholosist, and then the obtained knowledge was refined by sample cases. The previous works on developing expert systems usually rely on domain experts to provide all domain specific knowledge. But the method for acquiring knowledge directly from experts was inadequate in case of oriental medicine because it was hard to find an appropriate expert physician and the developing cost became too high. The motivation for this study came from the difficulties experienced with existing approaches which rely only on the expert or only on sample cases. In order to reflect the clinical knowledge contained in samples into the rough knowledge efficiently, the domain dependent strategies and heuristics were used in finding the subset of knowledge to be refined, generating new rules, and adjusting certainty values in rules. This method led to the development of Oriental Medicine Expert System (OMES). For the purpose of comparing the refinment abilities of OMES with the learning ability of the neural network, Oriental Medicine Neural Network was consturcted. These two systems have been compared with the system whose knowledge base was constructed by direct help of domain experts (OLDS). Among these systems, OMES was considered to be superior to other systems in terms of performances, development costs, and practicality. In this study, the development of OMES is presented and the performance of OMES in comparison with other similar system is described.