An Automated English Writing Scoring and Feedback System by Deep Semantic Analysis

This code implements a complete automated English writing scoring and feedback system based on deep semantic analysis. It consists of five core modules working together: the data processing module loads and preprocesses the ASAP-AES dataset; the deep semantic scoring model (BERT with multi-task learning and attention mechanism) predicts overall essay scores and four dimensional scores (linguistic expression, content quality, structural logic, grammar accuracy); the issue detection module identifies writing errors using pattern matching and BiLSTM-CRF; the feedback generation module combines template-based and generative approaches to produce hierarchical, personalized feedback.