Knowledge representation using structured modeling
Srikanth Chari, Arthur M. Geoffrion, E. Burton Swanson · University of California, Los Angeles eBooks · 1988
This dissertation's research studies the knowledge representation capabilities of Structured Modeling (Geoffrion, 1987). According to Structured Modeling (SM), a model is defined in terms of elements, each of which is either postulated as primitive or is defined in terms of other previously defined elements. Elements are organized into genera and genera into modules, allowing the user to view the model at various levels of detail. The research has two objectives: (a) to describe the core concepts of SM using predicate logic, and (b) to clarify the modeling capabilities of SM with respect to semantic data models. The first objective aims to provide a better understanding of SM building a bridge between SM and a well defined mathematical system. Inference rules and theorem proving techniques, applicable to models defined in logic, can be applied to models expressed in SM. The research discusses the correspondence between structured models and logic, while proposing a system using predicates, functions, and clauses (more specifically Horn clauses) to represent structured models. The research demonstrates how any structured model or model schema can be represented in logic, and defines necessary and sufficient conditions (in clausal form) under which a set of clauses represents a structured model and a model schema. It describes a PROLOG based implementation, developed to test the validity of the representation. The second objective, examining the modeling capabilities of SM, serves to demonstrate that SM is capable of representing the major abstractions of interest to data modelers. For this purpose, the research first formally defines the fundamental components of semantic data models. Providing formal definitions may be of independent interest in itself, since much of the literature on semantic data models defines the major components using examples rather than formal definitions. We discuss the representation of these components using the core concepts of SM and then examine the capabilities of Structured Modeling Language (SML) to express such information. We elaborate on the strengths and weaknesses of SM and SML as semantic data models and suggest some extensions to remove the limitations. (Abstract shortened with permission of author.)