Debugging Shape Definitions for use in Sketch Recognition

Why: To date, sketch recognition systems have been domain-specific, with the recognition details of the domain hard-coded into the system. Developing such a sketch interface is a substantial effort. We propose instead that recognition be performed by a single domain-independent recognition system that uses a domain specific sketch grammar (an approach used with some success in speech recognition [3]). Programmers could then create new sketch interfaces simply by writing a domain description sketch grammar describing all of the domain specific information. The domain description consists of several shape definitions which specify how a shape is drawn, displayed, and edited in a domain. Domain descriptions are written in LADDER, a language for describing how shapes are drawn, displayed, and edited in sketch recognition. Although efforts are being made to make LADDER as intuitive as possible, shape definitions can often be difficult to describe textually as it is much more intuitive to draw a shape rather than type out a verbal description. Thus we would like to build a user interface to aid sketch interface designers in writing descriptions. Often developers may forget a constraint, allowing unintended shapes to be recognized. Or developers may add too many constraints, and the developers intended shape may not be recognized. Thus we have built a graphical user interface help the developer test when a definition is over or under constrained. How: The graphical user interface will allow the developer to type a definition, open a previously typed definitions, or automatically generate a definition based on a single drawn example [2]. The developer will then be able to debug the shape definition, not only checking the syntax, but also checking if the definition is over or under constrained. Figure 1 shows an example of what the GUI may look like.