Graph Neural Network-Enhanced Knowledge Graph for Content Mining and Intelligent Decision-Making in Ideological and Political Courses

This article puts forward a graph neural network-strengthened content excavation and decision-making mechanism for ideological and political building in college curriculum. A four-layer different kind knowledge graph (course-knowledge point-ideology politics element-education target) is constructed from multi-source teaching texts by means of BERT-CRF, dependency analysis, and rule mode plates, and it is deposited in Neo4j. Teaching behavior data are put into graph nodes and made consistent with them to construct a semantic-behavior combined expression, on which a GNN carries out the study of node embeddings and an explainable scoring function hence supports course adaptive evaluation and resource recommendation. Experiments which are carried out on multiple courses show that this method has better performance than rule-based and structure-only baseline methods, and thus enhances the transparency of ideology and politics embedding work.