Adaptating the levenshtein distance to contextual spelling correction

摘要

© Technomathematics Research Foundation.In the last few years, computing environments for human learning have rapidly evolved due to the development of information and communication technologies. However, the use of information technology in automatic correction of spelling errors has become increasingly essential. In this context, we have developed a system for correcting spelling errors in the Arabic language based on language models and Levenshtein algorithm. The metric distance returned by the Levenshtein algorithm is often the same for multiple solutions in correcting a wrong word. To overcome this limitation we have added a weighting based on language models. This combination has helped us to screen and refine the results obtained in advance by the Levenshtein algorithm, and applied to the errors of Arabic words. The results are encouraging and demonstrate the value of this approach.