Programming Conceptual Models Using Conceptual Graph Modules

This article introduces some basic functions and architectural issues for building a tool for programming conceptual models for explainable expert systems. The programming tool is independent from a particular problem class or problem solving method. Our work is based on the KADS knowledge engineering methodology which results in a conceptual model, and we use Sowa ́s Conceptual Graphs (CG) as a knowledge representation formalism. KADS' domain knowledge is easily to map into CGs, whereas KADS must be modified, to enable declarative programming of its inference knowledge. We show that units of inference knowledge can be described by using modules of semantic networks, or by using Sowa's terminology, a canon. These inference knowledge modules are instantiated from generic descriptions. All semantic networks, which belong to a conceptual model, are part of a modular knowledge base, each module representing the knowledge on its own right level of granularity. Views connect the declarative representation of inference knowledge to parts of the domain knowledge. Some features for dealing with modules are added to CGs, without modifying their expressive power. We give an example, how a comceptual model could be represented using CG modules.