A Correction Model for Real-word Errors

Abstract

Spell Checker is used to identify and correct mistakes made by users while writing text and the mistakes are generally spelling mistakes. An intelligent spelling correction system SMC is proposed to automatically correct spelling mistakes in text-editor or text documents using contextual information of the confused words. The system is capable to correct words belonging to the set of confused words fed into it if they are contextually wrong. In this technique, an algorithm to identify and correct real-word errors is proposed. One phase of algorithm uses trigram approach to correct spelling mistakes and the other phase of algorithm uses Bayesian approach to correct spelling mistakes. Brown corpus is used as a training set and a set of commonly confused words is used in this case. Selection of words in other phase of algorithm uses synonyms derived from dictionary in the scenario when words are not found in the corpus. Comparative analysis of the proposed approach with tribayes has also been performed to identify the accuracy of SMC. The results indicate that SMC gives higher accuracy for spelling mistakes identification and correction for the commonly confused words as compared to other spelling correction algorithms.