<div style="text-align:justify;"> This book introduces an automatic grammatical error correction model on sentence-level based on deep learning. For the first time in error correction tasks, native Chinese corpus (literary works) and sentence classification information (acceptable or unacceptable) were incorporated into the Transformer model for training. Previous research often overlooked the significant impact of native language information on error correction. Experiments demonstrated a substantial improvement in model accuracy, increasing from 9.7% to 48% with the inclusion of native corpus and classification information. </div>