Abstract

An approximate word-matching algorithm for Chinese is presented. Based on this algorithm, an effective approach to Chinese spelling error detection and correction is implemented. With a word tri-gram language model, the optimal string is searched from all possible derivation of the input sentence using operations of character substitution, insertion, and deletion. Comparing the original sentence with optimal string, spelling error detection and correction is realized simultaneously.