David Roach

An interactive source commenter for Prolog programs

Prolog meta-circular interpreters, i.e., interpreters for Prolog written in Prolog, perform at least two operations on an object program - they parse it and execute its instructions. There is a useful variant of the meta-circular interpreter, the meta-circular parser, which as its name suggests, parses an object program without executing its instructions. The value of such a parser is that it provides an elegant means to modify Prolog source code. As the object program is parsed, new information in the form of additional instructions, comments, etc., can be selectively inserted. The Prolog source code commenter we describe is a meta-circular parser with facilities added to allow a user to interactively enter comments. As a Prolog program is parsed into its basic components, the user is allowed to view that component and enter an appropriate comment. The result is a new fully commented (and formatted) source program.

An expert system for determining government relocation allowances

Abstract We describe an expert system whose rule base is a formalization of the subset of the federal travel regulations on relocation allowances. Its purpose is to help government employees making a permanent change of station to identify the reimbursable expenses associated with that change of station. We discuss the structure of the regulations, the implications of this structure for the rule base and inference engine, and the components of the user interface.

Approximate string matching and the automation of word games

It is shown how approximate sting matching can be involved in the automation of various aspects of word game construction or solution. In the discussion, the authors try to identify and explicate the underlying issues in computational linguistics and suggest techniques which have been used to address these issues. It is shown that the two main issues which arise in this context have to do with lexical processing and heuristics, issues which arise in more practical contexts as well.<<ETX>>

Expert systems as overlapping logical theories

A model-theoretic characterization of the components of logic-based expert systems which employ pure metalevel inference regimes is discussed. The roles of the basic components of such systems in the underlying first-order theories are specified. This includes the domain-specific rules, metalevel inference engine, and intrinsic interpreter. Fragments of an actual expert system are used to motivate and illustrate the analysis.<<ETX>>

An alternative to constraint logic programming for managing domain arithmetics in Prolog expert systems

We present a method for including procedural arithmetic knowledge in a declarative rid-base. The declarative and procedural components am integrated in such a way that the procedural component can be represented naturally in the rulebase but stilI be an integral part of the proof tree generated by the inference engine. A way of organizing explanations based on the integrated proof tree is discussed. Prolog is used to illustrate the implementation of the integration strategy.

“Tuning” an ASM metric: a case study in metric ASM optimization

Wc present an approximate string matching case study. An optimized version of the edit distance algorithm is described which has proven more accurate for a particular commercial application than the existing (benchmark) algorithm. The cvoluhon and nature of the optimization are detailed and test results are presented.

A logic based approach to the recognition of overlapping 2-D geometrical objects

A prototype image recognition system that illustrates and offers a particular solution to the problem of relating the procedural and declarative aspects of the image recognition process is described for overlapping plane geometrical objects. A hybrid approach comprising both declarative and procedural components in which the respective roles are clearly distinguished is applied. The usual definitions of the target geometrical objects are represented by production rules, and an expert system inference engine is used to identify objects based on them. The low-level input is preprocessed by the procedural unit of the system in order to provide suitable input to the relatively high-level expert system.<<ETX>>

RAP: relocation allowance planner, a rule-based expert system with self-defining documentation features

Government employees who are relocating are often eligible for reimbursements for expenses incurred during the relocation process. However, a host of complex government regulations must be examined in order to determine which expenses, if any, are reimbursable. In response to this problem, one set of the regulations, Appendix B of the Health and Human Services (HSS) Travel Manua1[6], was encoded into an expert system called RAP: Relocation Allowance Planner[ 1,7]. Although the user interface of RAP is written in C, the rule base and inference engine of RAP is written in the fifth-generation logic programming language Prolog[3].