A visual knowledge representation language for layout problem

Abstract

A knowledge programming approach is proposed in order to solve a layout problem. This approach describes a problem in a two-dimensional language called a visual language, applies to the problem a form of the knowledge that is also described in the visual language, solves the problem, and represents the solution also in the visual language. The knowledge is represented by visual production rules. A visual production rule has conditional and action parts that are represented by the views written in the visual language. A view is composed of visual objects, visual predicates, and visual pattern operators. The reasoning algorithm employed in this approach is forward chaining that is based on the pattern matching represented in the visual language. The approach makes the description of the knowledge about layouts simpler than with conventional approaches. This provides greater benefits in knowledge acquisition. It also allows one to describe ambiguous knowledge about the layout of the target objects without detailed coding. This approach is of interest as a means of implementing an inference process that executes while leaving intact ambiguous information such as that which humans handle. >