A Knowledge-Based Approach to Support Analytic Query Answering in Semantic Data Lakes

The increased flexibility brought by Data Lake technologies, along with size and heterogeneity of quickly changing data sources, bring novel challenges to their management. Making sense of disparate data and supporting users to identify the most relevant sources for a given analytic request are indeed critical requirements to make data actionable. This is particularly relevant in data science applications, where users want to analyse statistical measures from a variety of data sources. To this aim, in the paper we introduce a knowledge-based approach for a Semantic Data Lake, capable of supporting efficient integration of data sources and their alignment to a Knowledge Graph representing indicators of interest, their mathematical formulas and dimensions of analysis. By leveraging manipulation of indicator formulas, a query-driven discovery approach is exploited to dynamically identify the sources, along with the needed transformations, to respond a given .