Abstract

Entity misalignment in knowledge graphs (KGs)—caused by noisy data or inconsistent naming—severely undermines the accuracy of academic collaborator recommendation systems. To address this, we propose a KG-based scholar recommendation system featuring a novel LLM-powered entity alignment method. Our-stage approach first identifies potential matches unsupervisedly, then leverages LLMs' semantic understanding for precise alignment. This high-fidelity alignment directly enhances KG quality, leading to more accurate recommendations. By resolving core entity ambiguity issues, our system aims to significantly improve recommendation reliability. Evaluation on real-world datasets will validate the effectiveness of the alignment method and its impact on recommendation performance.

Download

Share

COinS