Development of an Automatic English Writing Assessment System Based on Deep Learning

The development of an automatic assessment system for English writing (AES) is an important topic in the field of educational technology, but existing methods have significant shortcomings in scoring dimensions and long text processing. Therefore, this study proposes a multimodal fusion deep learning framework that achieves multidimensional scoring through hierarchical feature extraction and dynamic optimization strategies. The model first integrates the pre-trained language model with a graph neural network (GNN), then designs a multi-scale context-aware module that combines local grammar rules with global semantic coherence analysis. A dynamic weight allocation mechanism is introduced to adapt the scoring criteria to different types of questions and differences in student abilities. The experiment shows that this method is significantly better than existing benchmark models in terms of rating consistency and feedback practicality. The research results promote the intelligent transformation of the education evaluation system.