Presenter
Benjamin Riesenberg (@briesenb)
Abstract
The Oregon Digital online repository stores RDF metadata for more than 500,000 digital cultural objects. Creation of this metadata is enabled by configuration of a property IRI for each descriptive field, more than 50 fields requiring IRI values, and mechanisms to retrieve, store, and display human-readable labels for RDF resources from dozens of value vocabularies. These features represent years of effort by the Oregon Digital team as an early adopter of RDF for digital-collections metadata.
The repository enables collection-level RDF exports for administrators, but this data is not suitable as-is for publication due to a variety of issues. These include inadequate or incorrect data typing, lack of language and script tags, erroneous predicate IRIs differing from intended terms, the presence of technical and administrative metadata which doesn’t provide useful information for data reuse, and data-modelling issues.
This talk presents example usage of the SPARQL query language to address the issues above, transforming and preparing RDF exports for publication. This relies on SPARQL functions to construct IRIs and data-typed or language-tagged strings where needed, graph patterns to place values in a structure conforming to a desired data model, federated queries to integrate additional data, and the CONSTRUCT query form to yield output graphs. The example workflow utilizes the Python rdflib library’s implementation of the SPARQL 1.1 Query Language for processing.