Article: Large language model-driven student career knowledge graph and interpretable adaptive reasoning Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.70 pp.68 - 96 Abstract: In response to challenges in temporal modelling, heterogeneous data fusion, and recommendation transparency in student career development, this paper proposes a temporal knowledge graph method based on large language models and interpretable reasoning. The approach designs a dynamic graph with time intervals to capture skill evolution, builds a self-validating extraction pipeline to automatically extract temporal information from unstructured resumes, and integrates symbolic logic with vector matching for interpretable reasoning. Experiments on CareerHop demonstrate strong recommendation accuracy with area under the curve reaching 0.842, an 11.7% improvement over graph neural networks, and temporal extraction accuracy reaching 0.957, a 57% increase over rule-based baselines. This technical approach addresses limitations of static representations in capturing ability growth and provides an accurate, transparent solution for high-stakes career decisions. Inderscience Publishers - linking academia, business and industry through research

Title: Large language model-driven student career knowledge graph and interpretable adaptive reasoning

Authors: Tong Shen

Addresses: Puyang Institute of Technology, Henan University, Puyang, 457000, China

Abstract: In response to challenges in temporal modelling, heterogeneous data fusion, and recommendation transparency in student career development, this paper proposes a temporal knowledge graph method based on large language models and interpretable reasoning. The approach designs a dynamic graph with time intervals to capture skill evolution, builds a self-validating extraction pipeline to automatically extract temporal information from unstructured resumes, and integrates symbolic logic with vector matching for interpretable reasoning. Experiments on CareerHop demonstrate strong recommendation accuracy with area under the curve reaching 0.842, an 11.7% improvement over graph neural networks, and temporal extraction accuracy reaching 0.957, a 57% increase over rule-based baselines. This technical approach addresses limitations of static representations in capturing ability growth and provides an accurate, transparent solution for high-stakes career decisions.

Keywords: temporal knowledge graph; explainable recommendation; large language model; LLM; person-job matching.

DOI: 10.1504/IJICT.2026.154397

International Journal of Information and Communication Technology, 2026 Vol.27 No.70, pp.68 - 96

Received: 06 Mar 2026
Accepted: 03 Apr 2026
Published online: 25 Jun 2026 *

Title: Large language model-driven student career knowledge graph and interpretable adaptive reasoning

Keep up-to-date