The project implements an end-to-end deep learning model for Chinese Mandarin speech recognition based on gated convolutional networks similar to Facebook’s Wav2letter architecture. It uses Gated Linear Units (GLU) as activation functions for faster convergence. The model is trained on the AISHELL-1 dataset containing 150 hours of recordings covering over 4000 Chinese characters. Evaluation uses Character Error Rate (CER) as the metric, with the validation set reaching approximately 11% CER.