This paper deals with problems of advanced knowledge representation for case-based reasoning. Beyond classical attribute/value pair representations, there is a more sophisticated perspective of cases as structured objects. Structural knowledge representation and reasoning allows for an immediate transfer of the results of computing similarity to case adaptation, which is currently the most crucial part both in case-based reasoning theory and in applications. The concepts of structural similarity are motivated by the authors' experience in a large application-oriented research and development project. The key ideas and concepts underlying structural similarity concepts are introduced. Based on these conceptual preliminaries, there is developed some approach towards axiomatization of structural similarity. Similarity concepts are formalized as operators with basic properties which can be lucidly axiomatized. The authors derive a few results exhibiting the power and limitations of fundamental approaches to structural similarity.
Case-Based Reasoning, Structural Similarity