Research on the Construction of Knowledge Graphs for Text-Based English Teaching Mining Based on Deep Learning

Existing knowledge graph construction methods have deficiencies in semantic understanding and dynamic knowledge updating, and are difficult to adapt to complex teaching scenarios. This paper proposes an English knowledge graph construction algorithm (DSL-KG) based on dynamic semantic learning. Through deep semantic mining and dynamic association modeling, it realizes the automatic extraction and graph construction of knowledge points in teaching texts. The DSL-KG algorithm builds entity recognition and relationship classification models with BERT as the core, combines GNN to model the dynamic relationship between knowledge points, and adopts a multitask learning mechanism to achieve joint optimization. The experimental simulation is based on English textbooks and test corpora. The results show that the DSL-KG algorithm achieves an F1-score of 93.8% in entity recognition, an accuracy rate of 91.5% in relationship classification, an increase in knowledge graph coverage to 87.2%, and a construction time of 4.8 seconds per text, which shows significant advantages over traditional methods and existing deep learning models. Practical application verification shows that the algorithm can efficiently generate knowledge graphs, and provide knowledge point recommendations and learning path planning support in intelligent teaching systems, providing a technical foundation for the intelligent development of English teaching.