Deep Learning Models and Knowledge Graph Construction: Recommendations for English Course Learning

This paper briefly introduces the bidirectional encoder representations from transformers-bidirectional long short-term memory-conditional random field (BERT-BiLSTM-CRF) algorithm and the BERT-BiLSTM algorithm for constructing the knowledge graph of English course resources. The former is used to extract entities, and the latter is used to extract relations between entities. Then, a constructed knowledge graph was used to recommend English course resources to students. Finally, simulation experiments were conducted. In the experiments, the recognition performance of the BERT-BiLSTM-CRF and BERT-BiLSTM algorithms for entities and the relations between entities was verified by comparison with the traditional back-propagation neural network and convolutional neural network algorithms, and the performance of the knowledge graph-based recommendation algorithm proposed was verified by comparison with the graph neural network-, association rules-, and collaborative filtering-based recommendation algorithms. The results showed that the BERT-BiLSTM-CRF and BERT-BiLSTM algorithms could effectively identify entities and the relations between entities in English course resources. The knowledge graph of English course resources as a whole could be divided into three parts: syntax, morphology, and phonics. The recommendation algorithm based on a knowledge graph could recommend English course resources to students more effectively.