A Method for Integrating Heterogeneous Data into a Knowledge Graph

Christoph Göpfert, Sheeba Samuel, Martin Gaedke

Abstract: Knowledge graphs have gained popularity in many areas as a means of representing knowledge in a structured format. A large number of approaches for knowledge graph construction have been developed based on unstructured, semi-structured, or structured data sources. Many approaches focus on narrowly defined application areas or specific data source formats. In this paper, we describe a systematic method for knowledge graph construction from sources with heterogeneous data formats. The method comprises steps from ontology development to data acquisition and integration, mapping, and data refinement as well as the evolution of the constructed knowledge graph. We evaluate our approach using the case study of TUCgraph, a knowledge graph covering entities from the academic environment. This knowledge graph is actively being used within a research information system, demonstrating the practical applicability and effectiveness of our method.