A model and a fast algorithm for multiple errors spelling correction

Abstract

A new model for multiple errors spelling correction is proposed. The model handles insert, delete, change, and transpose errors. In the new model, we put constraints on possible editing sequences to reflect the error occurrence phenomenon in spelling, resulting in an error measure different from the traditional editing distance error measure. Properties of the “error distance matrix” between two character strings are studied under the assumptions of the new model. A cut-off criterion has been discovered, which can detect whether the error distance between two character strings is greater than a prespecified value during the calculation. Based on this cut-off criterion, a fast algorithm has been developed to find the nearest neighbors of a given character string in a dictionary. Experiments have been conducted with results showing that the cut-off criterion can greatly cut down the computation time needed for the nearest neighbor searching.