Constrained Attribute Grammars for Recognition of Multi-dimensional Objects

Handwriting recognition is now a standard feature in many hand-held computers. In most systems, recognition is currently limited to recognition of handwritten text and graphics. Ilowever, there is a need to extend recognition to multidimensional domains that are traditionally difficult to input with a keyboard on a desktop computer. In this paper, we address the problem of recognizing multidimensional objects by introducing a new class of grammars that we call constrained attribute grammars. In a constrained attribute grammar, semantic information is captured by attributes, while spatial relationships arc capture by constraints on the attribute values. In addition, the concepts of keyword and relevance of a keyword are considered to reduce the computational complexity of parsing such grammars. A computationally efficient parsing algorithm based on these concepts is also presented.