“Using Linked Data” can mean producing it or consuming it. For small sets of data, producing Linked Data could be as simple as assigning unique identifiers to records (see What is a URI?) and sharing them. Production at large scale would probably require either of the following options:
- Create a Linked Data database (also known as an RDF database or a triple store) where data can be inserted through forms directly in the database as Linked Data.
- Export the data from an existing database and convert it to Linked Data using a template and publish it on an RDF database. This can be done periodically if the underlying database is updated regularly. The template is built through a process called ‘mapping’ (sometimes it is called ‘translation’). The template takes the fields of the underlying database and ‘maps’ them to what we call an ontology (e.g. the CIDOC-CRM) which is a set of universally agreed abstract entities. For example, a database field which stores terms describing the material of objects (e.g. ‘paper’) is mapped to the CRM ‘E57 Material’ entity. The local database field is thus replaced by a universally agreed entity that everyone understands.
To consume Linked Data, one queries the RDF databases using the SPARQL query language which is a technical language. In practice, software can be produced to make such querying user friendly again through the use of forms.