Syntactic Error Detection System Using HMM

Having an error detection and correction system is a fundamental requirement for any word processing application such as MS Word, Applix Word, JWPce, KWord, etc. Despite various efforts to develop such systems using rule-based, statistical-based, and other machine learning approaches, none of them have been satisfactory. The author of this research proposes an algorithm that utilizes the Hidden Markov Model to detect grammatical errors in input sentences. The Viterby algorithm is used to implement the Hidden Markov Model, and an annotated corpus from ILCI is used to calculate the HMM parameters. The results of testing the system on three types of datasets showed an overall precision of 100%, recall of 93.83%, and an f-measure of 96.7. The proposed algorithm has the potential to be used in the development of similar systems for other Indian languages.