Assessing the Overlap of Science Knowledge Graphs: A Quantitative Analysis

Abstract

Science Knowledge Graphs (SKGs) have emerged as a means to represent and capture research outputs (papers, datasets, software, etc.) and their relationships in a machine-readable manner. However, different SKGs use different taxonomies, making it challenging to understand their overlaps, gaps and differences. In this paper, we propose a quantitative bottom-up analysis to assess the overlap between two SKGs, based on the type annotations of their instances. We implement our methodology by assessing the category overlap of 100,000 publications present both in OpenAlex and OpenAIRE. As a result, our approach produces an alignment of 71 categories and discusses the level of agreement between both KGs when annotating research artefacts.