Attila Varga from ANETI Lab, Corvinus Institute for Advanced Studies have developed an innovative method to determine the similarity among academic papers and authors by leveraging citation networks. This method, based on transition probabilities, streamlines the literature search process by avoiding the need for complex classification systems and clustering issues. It provides a precise, continuous metric of similarity and has proven more effective than conventional techniques such as Node2vec in capturing the overarching structure of various academic fields. The research, which includes various testing scenarios to refine this approach, also discusses the practical implementation of this similarity measurement, particularly in estimating research interest similarities among individual scientists. To facilitate wider use, the team has released a Python package that can compute all the metrics explored in the study.

The preprint is available on arXiv (LINK).

Varga, A., Kojaku, S., Nascimento Silva, F. (2024) Measuring Research Interest Similarity with Transition Probabilities. https://doi.org/10.48550/arXiv.2409.18240.