JCSE, vol. 20, no. 3, pp.131-141, September, 2026

DOI: http://dx.doi.org/10.5626/JCSE.2026.20.3.131

Personalized Learning Push Technology Based on Knowledge Graph and Improved BiLSTM-CRF Model

Jiaying Liu and Haochuan Jia
School of Accounting and Auditing, Jilin Business and Technology College, Changchun, China

Abstract: Traditional online education employs a uniform delivery model, neglecting learners’ personalized needs, resulting in low efficiency and poor quality. To solve the above defects, a personalized learning push framework integrating knowledge graph (KG) and deep learning is proposed by the research. The study first constructs the knowledge background using an improved bidirectional long short-term memory network-conditional random field (BiLSTM-CRF) model with a multihead attention mechanism to enhance semantic capture, and builds learner profiles by combining entropy weighting and quantization models to analyze learning styles. It then introduces an educational KG to mine knowledge point connections and combines a multi-head self-attention optimized extreme depth factorization machine to achieve personalized learning delivery. In entity extraction, the model’s average recall, precision, and F1-score are 0.923 (±0.011), 0.916 (±0.014), and 0.938 (±0.012), outperforming similar models. In learner knowledge point mastery analysis, the model accurately analyzes test item mastery, with a maximum mastery threshold deviation of 0.02. In personalized learning recommendation experiments, the model adapts to various user groups; for direct learners, average accuracy is 0.815, showing the best overall performance. The proposed technology demonstrates good application effects and provides precise technical support for personalized recommendations in online education.

Keyword: No keyword

Full Paper:   0 Downloads, 2 View