This article presents a method to raise the generic reaction data to synthetic reaction knowledge and the implementaion of retrosynthetic analysis on computer. Retrosynthetic analysis is a key step of synthesis design. We use strategic bond based approach in this work as it has the advantage that its logic is easier to implement. We propose a computing model-classification model of reaction knowledge. This model, which can relevantly describe the reaction knowledge, is defined by using 3 rules: Rule A-reaction type; Rule B- conditions of the reaction and Rule C-the circumstances where the reaction will not work. This computing model can extract the most important and the most essential information from a huge amount of reaction data and convert them into the knowledge that is easily processed on computer. It also includes the information about the full scope and limitations of each transform, accordingly the capability to extrapolate in parsing phase can be enhanced.

Key words: CHEMICAL REACTION, COMPUTERIZED SIMULATION