The paper presents a prototype knowledge acquisition system which has been implemented based on Ripple Down Rules. The main aim of our system is to capture the human skills of recognizing and differentiating a character from other similar characters. The user can identify such differences graphically and (if required) modify the Knowledge Base without the guidance of a Knowledge Engineer. The paper outlines the basic framework and presents initial experimental results. Further, it discusses ways of scaling up the approach to handle very large number of Chinese characters.