Retargetable on-line recognition of diagram notations

This thesis demonstrates that there are common tasks which are applicable to different diagram notations in the construction of on-line diagram recognition systems. It does this by describing the implementation of a retargetable kernel for on-line diagram recognition which can be used as the basis of recognition systems for a number of different diagram notations. The retargetable kernel accepts input from an electronic whiteboard, a data tablet, or a mouse.; The kernel is augmented with domain-specific information and algorithms to target it to various diagram notations. Three prototype recognition systems, for UML, Math, and Molecular diagrams, are implemented using the retargetable kernel. The kernel provides for the integration of domain-specific algorithms and knowledge in a structured manner. This knowledge is used to refine the recognition performed by the kernel or to perform recognition tasks specific to individual notations.; The use of on-line information in diagram recognition is also explored. This thesis also demonstrates some of the benefits of on-line information to the recognition process. Symbol segmentation and recognition can be enhanced using information on the ordering of strokes.; As well, this thesis introduces a model which describes the process of recognizing diagram notations. This model can be used to describe the aspects of recognition which are accomplished by existing algorithms. It is also useful in the identification of common tasks in the process of diagram recognition. Tasks which are common to the recognition of a number of different notations are candidates for inclusion in a retargetable kernel.