Order- n correction for regular languages

Abstract

A method is presented for calculating a string B , belonging to a given regular language L , which is “nearest” (in number of edit operations) to a given input string α . B is viewed as a reasonable “correction” for the possibly erroneous string α , where α was originally intended to be a string of L . The calculation of B by the method presented requires time proportional to | α |, the number of characters in α . The method should find applications in information retrieval, artificial intelligence, and spelling correction systems.