Extending R2RML with Support for RDF Collections and Containers to Generate MADS-RDF Datasets

It is a best practice to avoid the use of RDF collections and containers when publishing Linked Data, but sometimes vocabularies such as MADS-RDF prescribe these constructs. The Library of Trinity College Dublin is building a new asset management system backed by a relational database and wants to publish their metadata according to these vocabularies. We chose to use the W3C Recommendation R2RML to relate the database to RDF datasets, but R2RML unfortunately does not provide support for collections and containers. In this paper, we propose an extension to R2RML to address this problem. We support gathering collections and containers from different fields in a row of a (logical) table as well as across rows. We furthermore prescribe how the extended R2RML engine deals with named graphs in the RDF dataset as well as empty sets. Examples and our demonstration on a part of the Library’s database prove the feasibility of our approach.