Yazici, M.Baygin, M.Karabey Aksakalli, I.K.2026-03-262026-03-262025979833158990510.1109/IDAP68205.2025.112223562-s2.0-105025000118https://doi.org/10.1109/IDAP68205.2025.11222356https://hdl.handle.net/20.500.14901/3726This study proposes a two-stage approach for the automatic detection and correction of grammatical errors in English. In the first stage, incorrect sentences were identified using the Random Forest classifier, which showed the highest success among various machine learning algorithms such as Random Forest, Decision Tree, Multi-Layer Perceptron, and Linear Support Vector Machines. In the second stage, grammatical correction was performed on only these sentences using the transformer-based BART model and T5 model. The dataset utilized is the Pranav GEC dataset published on Kaggle. After preprocessing, labeling, and balancing, sentence representations were generated using the TF-IDF vectorization technique. Experimental results indicate that the Random Forest model achieved a high F1 score in the classification task, while the BART model provided highly effective corrections, especially in terms of precision. The proposed method reduces computational cost while improving overall accuracy and provides more balanced results compared to traditional seq2seq-based approaches. This study offers a technically and pedagogically effective solution model for grammatical error correction systems and has practical potential for integration with language learning technologies. It also provides a foundation for future research in areas such as multilingual support, user feedback-based learning, and contextsensitive correction. © 2025 IEEE.trinfo:eu-repo/semantics/closedAccessBartDeep LearningGrammatical Error CorrectionNatural Language ProcessingRandom ForestConference Object