Generalization Handling in a Dynamic Case Memory

In the present paper, we deal with the problem of the recovery from bad generalization in a dynamic memory of cases. The memory model we adopt is based on the EMOP structures and has been applied at the case-based component of a system integrating case-based reasoning and prototypical knowledge. Two different types of bad generalizations (undergeneralizations and overgeneralizations) are dealt with by monitoring each norm of an EMOP and logically deleting bad generalized information. The second point leads to a conservative approach (justified by the use of a particular integrated architecture) which has the advantage that the recovery process can use all previously stored information.