An object-centred interpretation of the blackboard architecture for knowledge-based programming

Abstract

Progress with knowledge-based approaches to the production of reliable software has been slow but intense over the last decade. From individual efforts, environments have emerged to support these investigations with tools for knowledge representation and search. However, this support technology threatens to be overtaken as investigators, largely inspired by cognitive models, prefer to work with multiple representations of systems. The blackboard architecture offers a number of principles for reasoning amongst many representations of systems and might provide a framework in which knowledge-based programming could be supported. A suitable interpretation of the architecture must offer transparency to complex reasoning processes and include a high-level language for describing knowledge. In short the architecture must be made accessible for applications in the domain of programming. This thesis describes an object-centred interpretation of the blackboard architecture for knowledge-based programming. Solutions are developed on blackboards: described as assemblages of objects. Programming knowledge for particular applications and strategies for applying that knowledge are represented as rules in a high-level language known as APPEAL. Rules are themselves objects and occupy the blackboard. This interpretation, embodied in the ENCORES system, brings transparency with a previously unattainable flexibility for engineering programming knowledge.