Finding the neighborhood of a query in a dictionary

Abstract

Many applications require the retrieval of all words from a fixed dictionary D, that are 'close' to some input string. The paper defines a theoretical framework to study the performance of algorithms for this problem, and provides a basic algorithmic approach. It is shown that a certain class of algorithms, D-oblivious algorithms, can not be optimal both in space and time. This is done by proving a lower bound on the tradeoff between the space and time complexities of D-oblivious algorithms. Several algorithms for this problem are presented, and their performance is compared to that of Ispell, the standard speller of Unix. On the Webster English dictionary the algorithms are shown to be faster than 'Ispell' by a significant factor, while incurring only a small cost in space.<>