Ripple-down rules with censored production rules

Ripple-Down Rules (RDR) has been successfully used to implement incremental knowledge acquisition systems. Its success largely depends on the organisation of rules, and less attention has been paid to its knowledge repre- sentation scheme. Most RDR used standard production rules and exception rules. With sequential processing, RDR acquires exception rules for a particular rule only after the rule wrongly classifies cases. We propose censored produc- tion rules (CPR), to be used for acquiring exceptions when a new rule is created using censor conditions. This approach is useful when we have a large number of validation cases at hand. We discuss inference and knowledge acquisition al- gorithms and related issues. The approach can be combined with machine learn- ing techniques to acquire censor conditions.