Definition of mapping types / scenarios

[simontaurus]

@linikujp, @matentzn, @sveinugu, @jesper-friis and others

I was following the resources of this WG with interest - also try to join the regular meeting soon.
Creating proper definitions of the broad term "mapping" seems a very important step.
For this need to cover the following cases which I only see this partially covered by the current discussion:

• non-LD to LD mappings, usually conventional structured data (CSV, JSON, XML) to RDF, somehow reflected by Schema-Mapping but a schema is not present in every case (e.g. simple JSON document) and LD-Schemas like OWL should be excluded here
• LD to LD, usually RDFS to RDFS (including OWL to OWL) which is covered by Ontology-Mapping (Classes, Properties) and Entity-Mapping (Individuals)
• non-LD to non-LD mapping, which should be composed be non-LD to LD, LD to LD and LD to non-LD. This is what most people in the end are looking for, LD/RDF as an invisible bright that magically make application interoperable (see also JSON-LD for non-LD to LD and LD to LD mappings #85)

[matentzn]

@simontaurus thanks for your question and good to meet you at the last call.

We currently zone in on two families of mappings for this WG:

1. Entity mappings are "correspondences between named entities" (or more precisely, symbols representing enties, like database identifiers or linked data resources).
2. Schema mappings are very vaguely correspondences between structures. This captures anything from an RML specification (CSV/SQL to RDF) and a classic schema mapping (PROV -> DCAT) to an ETL script that transform a dataset (even one without an explicit schema) to another (again, possibly without a formal schema). In the last case, the schema is still there, just implicit - that's how we justify squeezing this under "schema mapping". I am sure there will be people out there severely offended by this choice, but we are about defining a solid metadata spec, not to becoming the worlds authority on what a schema is :P

Your three cases fall under (2), but your LD-LD examples sound like our case (1) - however, I would have rather not said "entity mapping" as and example of LD-LD, but maybe something else like... SPARQL construct? entity mappings, class class, etc, are not really LD specific.

See slides for more (actually check the whole deck if you are interested).