Knowledge Graph-Based Query Rewriting in a Relational Data Harmonization Framework

There are diverse data providers, storage formats and data schemas in many modern application domains from sales and marketing to health care. This leads to a high demand for a harmonized data management platform that hides the heterogeneity of the system from the end user. Using an abstraction layer that provides a single logical view of all data located in disparate data sources, makes the implementation of a harmonized data management platform effortless. This abstraction layer is called Data Virtualization and it can query the data sources via a single query which is usually in the common SQL format. However, the user needs to know where the desired data is located before submitting the SQL query to the virtualization middleware. In this paper, we present how to capture inter-database associations of relational data stores as a Resource Description Framework (RDF) graph for the purpose of automation in Data Virtualization. Furthermore, we propose an approach to enrich a naïve input SPARQL which can query the RDF graph and translate it to a SQL query that will be fed into the virtualization layer. Combining these two approaches results in the transparency of the data harmonization framework.