HCPRVR: An Interpreter for Logic Programs

An overview of a logic program interpreter written in Lisp is presented. The interpreter is a Horn clause-based theorem prover augmented by Lisp functions attached to some predicate names. Its application to natural language processing is discussed. The theory of operation is explained, including the high level organization of the PROVE function and an efficient version of unification. The paper concludes with comments on the overall efficiency of the interpreter. An axiom is either an atomic formula, which can be referred to as a fact, or an expression of the form ( < .*. ) where both the conclusion and the premisses are atomic formulas. The symbol "<" is intended to be a left-pointing arrow. An atomic formula is an arbitrary Lisp ___-expression beginning with a Lisp atom. That atom is referred to as a relation or predicate name. Some of the other atoms in the expression may be designated as variables by a flag on their property lists. I INTRODUCTION III CALLING LOGIC PROGRAMS-HCPRVR, a Horn Clause theorem PRoVeR, is a Lisp program that-interprets a-simple logical formalism as a programming language. It has been used for over a year now at the University of Texas at Austin to write natural language processing systems. Like Kowalski [II, we find that programming in logic is an efficient way to write programs that are easy to comprehend. Although we now have an interpreter/compiler for the logic programming language Prolog [2], we continue to use HCPRVR because it allows us to remain in a Lisp environment where there is greater flexibility and a more familiar notation. This paper outlines how HCPRVR works to provide logic programming in a Lisp environment. The syntax of logic programs is given, followed by a description of how such programs are invoked. Then attachment of Lisp functions to predicates is explained. Our approach to processing natural language in logic programs is outlined briefly. The operation of HCPRVR is presented by giving details of the PROVE and MATCH functions. The paper closes with some remarks on efficiency.