Integrated Model to Develop Grammar Checker for Afaan Oromo using Morphological Analysis: A Rule-based Approach

Abstract

This study has implemented a rule-based approach on grammar checkers by integrating a spell-checker with a morphological analyzer to improve the Afaan Oromo grammar checker. A corpus containing about 300,000 words has been prepared to be used for spell-checker. About 300 grammar rules are constructed to detect the grammar error within the Afaan Oromo text and to suggest the possible grammar correction. The developed frameworks have experimented on the document having pairs of 100 correct and incorrect sentences. The experimental result for checking the spelling errors has scored 73% of recall, 76% precision, and 75% of F-measure. The score for suggesting the correct spelling is 78% of recall, 62% precision, and 70% precision F-measure while the evaluation result for detecting the grammar errors has 47% recall, 90% precision and 68% f-measure score. For suggesting the possible correct grammar on the detected error, the system has scored 61% recall, 71% precision and 66% f-measure. The overall performance of the developed system has a good performance. However, there is still a need to conduct further research to improve the Afaan Oromo grammar checker.