Much design knowledge is expressed and communicated in graphical form, and yet no practical method exists for automatically extracting this knowledge for use within computer-aided design systems. In addition, no methods exist for automatically comparing graphical designs for their similarities and differences. A framework in which these two issues (graphical knowledge extraction and comparison) can be addressed is created. One approach to extraction of graphical knowledge is the creation of an expert system which can employ not only facts and rules but can also reason about spatial patterns. The theoretical framework for such a prototypical design tool will be created. Once developed, this framework will be useful for all areas of design in which knowledge is expressed graphically. The working system that will be created under this grant will be a computer-aided architectural design (CAAD) system. Once the framework for extracting graphical knowledge from a drawing exists, the principal investigator will create an expert system for CAAD that will acquire graphical knowledge from a design data base. From a large set of designs, the expert system will make inferences about the characteristics of these designs.
