Article: Knowledge graph construction for online courses using enhanced BERT and BiLSTM Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.39 pp.82 - 106 Abstract: With the diversification of internet-based education and the need for cross-domain knowledge integration, course knowledge graph construction faces higher performance requirements. Existing methods often rely on manual labelling, perform poorly on low-frequency data, and fail to model nonlinear semantic relationships or real-time updates. To address these issues, this paper proposes a course knowledge graph construction model combining enhanced BERT with a bidirectional long short-term memory (BiLSTM) network. The model jointly captures semantic representations and sequential dependencies to build structured knowledge graphs. Experimental results show that the proposed method achieves a maximum relation prediction accuracy of 99.0%, with peak memory usage of 695 MB and a maximum response time of 79 ms, outperforming baseline models. In practical applications, the question-answering accuracy ranges from 90.2% to 95.4%. These results demonstrate that the model significantly improves accuracy, efficiency, and real-time adaptability, providing effective support for intelligent course knowledge graph construction. Inderscience Publishers - linking academia, business and industry through research

Title: Knowledge graph construction for online courses using enhanced BERT and BiLSTM

Authors: Jing Li

Addresses: College of General Education, Wuhan Business University, Wuhan, 430056, China

Abstract: With the diversification of internet-based education and the need for cross-domain knowledge integration, course knowledge graph construction faces higher performance requirements. Existing methods often rely on manual labelling, perform poorly on low-frequency data, and fail to model nonlinear semantic relationships or real-time updates. To address these issues, this paper proposes a course knowledge graph construction model combining enhanced BERT with a bidirectional long short-term memory (BiLSTM) network. The model jointly captures semantic representations and sequential dependencies to build structured knowledge graphs. Experimental results show that the proposed method achieves a maximum relation prediction accuracy of 99.0%, with peak memory usage of 695 MB and a maximum response time of 79 ms, outperforming baseline models. In practical applications, the question-answering accuracy ranges from 90.2% to 95.4%. These results demonstrate that the model significantly improves accuracy, efficiency, and real-time adaptability, providing effective support for intelligent course knowledge graph construction.

Keywords: bidirectional encoder representations from transformers; BERT; bidirectional long short-term memory network; internet-based courses; knowledge graph; TextRank.

DOI: 10.1504/IJICT.2026.153310

International Journal of Information and Communication Technology, 2026 Vol.27 No.39, pp.82 - 106

Received: 04 Jan 2026
Accepted: 04 Feb 2026
Published online: 01 May 2026 *

Title: Knowledge graph construction for online courses using enhanced BERT and BiLSTM

Keep up-to-date