Invited Talk : Tabular Data to RDF and Data Shapes

To enable analyses that span multiple sources of tabular data, eg for research in agriculture, researchers can start by associating their variables with URIs for properties from established authorities such as Dublin Core or from the more domain-specific NALT Concept Space of the USDA National Agricultural Library ("reconciliation"). Datasets are then modeled as sets of interrelated "data shapes", each with the properties used for describing specific types of entity, such as "cotton fiber samples". Models are edited in a straightforward spreadsheet format, Dublin Core Tabular Application Profile (DCTAP). The DCTAP serves as the source for automatically generating UML diagrams, HTML documentation pages, and machine-actionable validation schemas (ShEx) and as a target form for extracting data from tabular formats into RDF ("semanticization"). The resulting RDF knowledge graph is validated by ShEx; queried with SPARQL; and used for visualizations. The properties used in data shapes are added to SKOS collections ("property sets") in NALT. These growing property sets serve as starting points for creating future data shapes for future requirements.