Research on Intelligent Teaching Resource Recommendation System Based on Knowledge

In order to overcome the shortcomings of current recommendation systems such as deep semantic association and personalized recommendation, an intelligent teaching resource recommendation system based on knowledge graph is proposed. A semantic adaptive fusion model (SAFM) is proposed in this article. The model comprehensively utilizes the text features of teaching resources, knowledge graph embedding based on graph neural network and user behavior sequence information, and realizes dynamic feature fusion through adaptive weight mechanism, so as to more accurately capture the deep semantic relationship between resources and user personalized needs. The system architecture is divided into four levels: data collection, preprocessing, model training and user feedback. The data sources include textbooks, videos, exercises and domain knowledge annotated by experts. This experiment was conducted using data set from an online educational platform. Results show that compared with traditional collaborative filtering and content-based recommendation methods, SAFM model improves recommendation accuracy by 20 %, improves recall rate by 15 %, and improves F1 by 18 %. This system has an average response time less than 2 seconds. It can satisfy the requirement of real-time recommendation.