Title:Academic Literature Recommendation in Large-scale Citation Networks Enhanced by Large Language Models

Abstract:Literature recommendation is essential for researchers to find relevant articles in an ever-growing academic field. However, traditional methods often struggle due to data limitations and methodological challenges. In this work, we construct a large citation network and propose a hybrid recommendation framework for scientific article recommendation. Specifically, the citation network contains 190,381 articles from 70 journals, covering statistics, econometrics, and computer science, spanning from 1981 to 2022. The recommendation mechanism integrates network-based citation patterns with content-based semantic similarities. To enhance content-based recommendations, we employ text-embedding-3-small model of OpenAI to generate an embedding vector for the abstract of each article. The model has two key advantages: computational efficiency and embedding stability during incremental updates, which is crucial for handling dynamic academic databases. Additionally, the recommendation mechanism is designed to allow users to adjust weights according to their preferences, providing flexibility and personalization. Extensive experiments have been conducted to verify the effectiveness of our approach. In summary, our work not only provides a complete data system for building and analyzing citation networks, but also introduces a practical recommendation method that helps researchers navigate the growing volume of academic literature, making it easier to find the most relevant and influential articles in the era of information overload.