<p>Knowledge graphs (KGs) provide a graph-based data model for knowledge representation. A promising research direction refers to the computation of centrality measurements. Such measurements are usually applied to KGs to determine the relevance of their nodes and to characterize relevant relation-based phenomena occurring on those networks. However, those investigations do not ensure the validity of centrality measurements in KG-based analyses nor account for knowledge graphs’ edge types when computing such measurements. Centrality analyses that account for KG edge types can provide distinct insightful interpretations regarding the relevance of nodes, conditioned by the semantics encoded by the relationship types. This article demonstrates that centrality measurements apply to KGs, as they share properties with real-world complex networks. Second, we investigate how to support knowledge graph analyses exploring centrality measurements. To this end, we introduce novel knowledge graph-informed centrality measurements in which edge types are considered. We conduct a comparative study based on KGs generated from corpora of unstructured texts obtained from scientific documents. In particular, we explore texts from papers of the Semantic Web conferences domain, comparing conventional centrality measurements with KG-informed measurements. We seek to determine the effectiveness of both types of centrality measurements in encoding the relevance of entities represented by the KG nodes. Our results demonstrate such effectiveness, and we identify scenarios for which we can leverage both the conventional and KG-informed centrality measurements.</p>