A Correcting Model Based on Tribayes for Real-Word Errors in English Essays

Abstract

This paper addresses the problem of real-word spelling errors, and also the problem of omission of effective features due to deficiency of training set in spelling correction. then a method called RCW (real-word correction with Word Net) based on Tribayes is introduced, and it solves these problems to a certain extent. Drawing upon the context information, the score of ambiguous words are calculated and regarded as decisive factor for real-word errors correction in RCW. Moreover, the synonyms of the effective features ignored are extracted from Word Net, and we use them as feature so as to improve the accuracy of real-word errors correction. Experiment shows that RCW is able to provide a better performance than Microsoft Word 2007 on real-word errors correction.