Dale E. Fish and Robert McCartney
We are exploring the possibility of using case-based reasoning as an approach to diagrammatic reasoning. This paper describes a work in progress on a system that 'understands’ a diagram of a problem situation by finding correspondences between it and similar diagrams in its case memory. The input to the system is a diagram of the problem situation consisting of some number of simple elements such as circles and lines. The goal of the system is to determine the significance of the diagram and its constituent elements by recognizing the structure of the diagram and structures within it. The approach is to use high level perception (recognition of instances of categories and relations) to build a representation of the problem situation (a diagram), and infer additional information by finding correspondences between this representation and stored cases. The output is the problem situation diagram annotated with descriptions of the elements and relationships between them and any inferences the system may be able to make along with their justifications.