Development of a Hybrid AI-Based Tigrigna Grammar Checking and Error Correction System with an Interactive Graphical User Interface

Abstract

Tigrigna is one of the morphologically and syntactically rich but low-resource languages with limited annotated datasets, linguistic resources, and computational resources. Due to these challenges, developing an automated grammar analysis and correction is hindered. Therefore, this research presents the development of an automated Tgrigna grammar analyzer and correction system using a combination of rule-based, machine learning, and hybrid approaches using the Python programming language. The model is trained with a total of 2000 sentences, including 500 correct sentences and 500 incorrect sentences per class of agreement fault, gender biased fault, and tense fault sourced from books, online articles, and academic materials. In addition to this, the model was trained with word tokens of 625 nouns, 545 verbs, 440 adjectives, 14 pronouns, and 90 noun-prepositions and evaluated with 80% for training and 20% for testing. Experimental results indicate that the hybrid approach achieves an accuracy of 94%, significantly outperforming the pure machine learning model with an accuracy of 92% for random forest and naive Bayes, 91% for logistic regression, and 90% for linear svm and the rule-based model with an accuracy of 81%.  The model also integrated features of a graphical user interface (GUI) that enable it to identify errors and provide automated grammatical suggestions and corrections.