Design of University English Grammar Error Correction System based on Neural Network Algorithm

Abstract

This paper describes the design and assessment of a University English course focused on the writing process. A Grammar Error Correction System was developed based on Neural Network Algorithms and also uses state-of-art machine learning techniques that are, in this case, Transformer-based models like BERT and GPT. Notably, it can recognize and correct a multitude of grammatical errors in the text from English. Experimental results demonstrate an extremely high level of precision, recall and F1-score values that indicate the system's correct identification and correction of grammar errors. The usability and educational value of the tool were validated through user studies where participants appreciated its intuitive interface and helpful explanations. Findings underscore that Automated Grammar Correction tools may become the gist of language learning and teaching processes.