Design and Development of a Deep Ghost Gated Learning Framework For English Auxiliary Teaching System

Traditional English language instruction in colleges often relies on generalized curricula and static teaching tools, which do not cater to the varied learning paces and proficiency levels of students. This lack of personalization can hinder student engagement and limit the overall effectiveness of language acquisition. Furthermore, conventional systems fall short in leveraging data to adapt teaching strategies dynamically. To resolve these concerns, this study presents the design and implementation of a deep learning-based auxiliary teaching system for College English, powered by a Deep Ghosted Gated Recurrent Unit (DGDGGRU) network. The DGDGGRU architecture, known for its efficiency in modelling sequential data with fewer parameters than LSTM, enables the system to learn temporal patterns in student behavior and performance. The model is trained on the Learning Analytics Dataset for English Language Teaching (ELT), which contains student engagement logs, test scores, interaction records, and background information. This dataset is publicly available at Kaggle – Learning Analytics ELT Dataset. Experimentation outcomes show that the proposed DGGRU-based system achieved 98.8% accuracy, 97.5% precision, 97.9% recall, 97.7% F1-score, and an AUC-ROC of 0.99, outperforming conventional models. These findings highlight the potential of DGGRU in building adaptive, intelligent educational tools that enhance English learning outcomes and provide instructors with actionable insights.