This project is about grammatical error correction, and the focus here is article (“a”, “an”, “the”) errors. We use classification approach. And the experiment is 10-fold cross validation on NUCLE dataset. The precision, recall, and f1 value are 12.8%, 34.7%, and 0.19.