Grammatical error correction (GEC) deals with automatically correcting errors in written texts, including grammatical, spelling, punctuation, and word choice errors. Current GEC systems rely heavily on annotated corpora for learning, where human teachers have manually identified and corrected the errors in texts written by learners. Since manually preparing the training corpora is laborious and time-consuming, we will explore semi-supervised or unsupervised learning approaches to reduce this dependence. We will also explore generating more helpful and customized feedback to enable learners to more effectively learn to improve their writing.