Discovering Relationships Between Heterogeneous Declarative Mappings for RDF Knowledge Graph

Nowadays, Knowledge Graphs (KGs) are extensively used in companies, they are created using different techniques, mapping languages among them, and involve various designers teams. Mappings languages have been proposed to explicitly define mappings assertions between heterogeneous (semi)-structured datasets and the KG. The wide variety of mapping languages and their associated systems and the lack of a generic mapping language that covers all the existing languages, lead each expert/design team to use the mapping language they master and that fits their requirements, to enrich the company’s KG. This situation creates new sources of heterogeneous mapping assertions. A straightforward analysis of these mappings is very complex because of the strong differences of their syntaxes. In this article, we propose an approach to detect relationships between heterogeneous mappings from the analysis of their generated graphs, following a reverse ETL approach. This analysis is relevant for analyzing and maintaining the mappings, and may provide relevant insights in the datasets usage and design patterns defined within the global KG project, where the mappings play a central role. A set of experiments is provided to show the feasibility of the approach.