Increasing the efficiency of Prolog Lexical databases with N-gram Boolean cubes

Abstract

PROLOG has been shown to be an effective tool for expressing the logic of many problems dealing with parsing, natural language processing, and spelling verification [1,7,8,9,12]. As a class, these problems deal with the manipulation of lexical databases as Horn clauses. Since PROLOG does not generally differentiate between program clauses and data clauses, the internal representation and manipulation of data may not be optimal for a particular application. This paper discusses an alternative method of representing and manipulating lexical databases through the use of N-gram analysis, prefiltering, and integration with another high level language.