A Real-Time Recognition System for Handwritten Mathematics: Structural Development

This paper describes a scalable structural development process, which is part of a real-time recognition approach for handwritten mathematical input. The incremental development of the expression tree is important for real-time recognition, due to the requirement of instantaneous user feedback. Traditional parsing techniques are not suitable for this approach. However, the scalability of a grammar-based approach is appealing and has been incorporated into the structural development technique. The tree is developed, using the relationships discovered by detailed spatial analysis techniques. Due to the fact that unordered and multi-stroke symbols can affect previous increments of the expression tree, the system employs a backtracking technique to revert back to previous states. While this avoids the need to impose symbol ordering constraints on the user and reduces the complexity of relationship discovery, we will describe here how it also simplifies the structural development process.