Automatic generation of rules by transformation:LIFT

Abstract

Knowledge is the basic component of expert systems. However, in the rule-based systems, it is difficult for experts to express their expertise as rule, because the if-then-clause is not natural representation to experts. Therefore, it is necessary to provide more natural representation to the experts so that they can reveal their expertise more effectively and efficiently. For this purpose, we propose a graphical representation called symbolic diagram. From our experience, the experts (accountants in this case) could reveal their knowledge more efficiently by using this symbolic representation. To transform the symbolic diagram into rules automatically, we have developed LIFT (Logical Information Fast Transformer). The LIFT has such a generality to adjust to different symbols, rule syntaxes, and transformation strategies. A prototype of LIFT is developed using pascal on the micro computer, and is applied in constructing knowledge base of tax consulting system which uses SKI 2 developed by joint work of KAIST and Ssangyoung Software \& Data Co., Ltd.