User browsing information contains user’s interest preference, which is of great value for realizing personalized push of teaching resources. In this paper, we developed a user intent classification model grounded in IndRNN-Attention, extracted user intent from the user’s historical browsing records, and subsequently designed a network for explicitly modeling user intent based on knowledge graph (KG-UIN), and assessed the practical effect of collaborative filtering recommendation algorithms based on KG-UIN on the task of delivering teaching resources for higher vocational colleges and universities. The study reveals that the intent classification accuracy of the IndRNN-Attention model across the training set, validation set, and test set attains 96.97%, 80.44%, and 83.32%, respectively, all of which surpass those of the LSTM and the IndRNN model, which shows that the IndRNN-Attention model introduced in this paper yields a pronounced effect. In addition, the recommendation system based on KG-UIN achieves better results in user satisfaction, accuracy and surprise of recommended resources, which can be applied to the personalized push task of teaching resources in higher vocational institutions and universities.
