Opinion Leader Identification for Artificial Intelligence Domain Analysis Using a Graph-Based Model

Abstract

This paper addresses the challenge of navigating the rapidly evolving field of artificial intelligence (AI), using large language models as a representative example. It proposes a graph-based representation of the scientific community as an analytical tool for describing the structure of relationships between researchers and identifying research groups. The study also introduces an approach for detecting key figures and opinion leaders within the field. The underlying assumption is that analyzing the publications of such groups can help capture emerging trends in a timely manner and support informed decisions regarding the adoption and implementation of relevant technologies. Using this approach, a graph model was constructed based on open scientometric data: researchers are represented as nodes with additional attributes, while their relationships are encoded as edges. The influence of individual authors was quantified using PageRank centrality, and latent research groups were identified through the Louvain clustering algorithm. The results support the initial hypotheses: scholars with high PageRank scores are indeed recognized industry leaders, and the algorithm consistently identifies five clusters corresponding to real research and corporate structures. Overall, the proposed graph model can be considered a supporting tool for analytical characterization of the current AI research landscape and for monitoring emerging scientific trends.