Efficient generation of lexically proper palindromes

Abstract

Palindromes are strings of symbols which are symmetrical about the center. This paper outlines a method for generating certain types of palindromes, called lexical palindromes, which consist of legitimate English words. The method reported provides substantial pruning of a Prolog search tree by calculating the number of success nodes along certain search paths instead of visiting them, indexing words to improve database performance, and continuous analysis of current states to eliminate non-productive search paths. These efficiency measures allow lexical palindromes to be generated using a microcomputer. Our ultimate objective is to be able to estimate the ratio of sentential palindromes for any given length to the set of syntactically well-formed English sentences of that length with respect to a given lexicon and grammatical model. We shall take up the problem of determining the ratio of sentential palindromes to well-formed sentences in a sequel to this paper.