The Use of Natural Language Processing (NLP) in the Automated Detection and Correction of Grammatical Errors in Students’ English Writing

Abstract

Accurate, timely feedback on grammatical accuracy remains one of the most persistent challenges in English language teaching, particularly in large classes where teacher marking capacity is limited. This article reviews recent developments in the use of natural language processing (NLP) for the automated detection and correction of grammatical errors in students’ English writing. A narrative synthesis of 26 peer-reviewed studies and preprints was combined with an illustrative case study in which a hybrid Transformer–large language model (LLM) pipeline was applied to 240 essays written by EFL undergraduates. The review shows a clear technological trajectory from rule-based and statistical correctors through sequence-to-sequence neural models to pretrained encoders and, most recently, instruction-tuned LLMs capable of generating explanatory feedback. Reported systems now routinely exceed 80% precision on high-frequency error categories such as article and determiner use, although word-order errors and context-dependent preposition choice remain difficult. Pedagogically, automated written corrective feedback (AWCF) improves accuracy and revision behaviour, but over-reliance, uneven coverage of low-frequency errors, and hallucinated corrections by generative models require careful instructional design. The article concludes with implications for integrating NLP tools as formative, teacher-in-the-loop aids rather than stand-alone assessors.