Application of information retrieval system based on knowledge graph in education

To solve the problem that the information retrieval system in advanced higher mathematics teaching cannot deeply analyze the logical relationship of complex concepts, theorems, and formulas, which makes it difficult for users to understand the knowledge system fully, this article constructs a knowledge graph and designs a new information retrieval system to improve the integration efficiency of advanced higher mathematics teaching resources. This article uses ontology to semantically model the core concepts of advanced higher mathematics and define their attributes and associations. The bidirectional encoder representations from transformers (BERT) model is fine-tuned to adapt to advanced higher mathematics, automatically identifying and extracting semantic relationships between concepts and constructing triples. A semantic matching mechanism that integrates graph embedding is designed, combining BERT and translation embedding (TransE) technology to improve the degree of matching between queries and knowledge nodes. The comprehensive scoring sorting optimization algorithm is used to integrate semantic similarity, node weight, and structural similarity for sorting optimization. The graph neural network (GNN) algorithm is applied to dynamically generate paths based on the knowledge graph and support inference logic. Experiments show that the average response time of the knowledge graph-based advanced higher mathematics information retrieval system is 70.28 milliseconds; the accuracy of simple and complex queries is 95.61% and 86.42%, respectively; the coverage of associated resources is 95.5%; the average score of teaching adaptability is above 9 points. The research proves that the system in this article can integrate educational resources, enhance students’ understanding, improve teaching quality, and promote educational informatization.