An Information Geometry Approach to Analyzing Topic Evolution in Scientific Networks: From Physics to International Relations

This study presents a novel methodology for analyzing the evolution of scientific topics through the geometric framework of information spaces. Using mutual entropy-based distance metrics, the approach reveals dynamic relationships between scientific concepts over time, surpassing the capabilities of traditional keyword-based analyses. The framework quantifies the creative influence of publications linked to knowledge brokers by measuring the relative compression these agents induce on the geometry of knowledge networks. Applied to topics derived from ArXiv and JSTOR datasets, the methodology identifies patterns of topic evolution and evaluates the impact of key agents, such as publishers, journals, and countries. The findings offer actionable insights for strategic planning by academic journals, funding agencies, and research institutions, facilitating data-driven decision making to promote emerging research trends and interdisciplinary collaborations.