DDI-CDI & JSON-LD

DDI-CDI Dagstuhl Seminar – 19-24 Sept. 2021

Pierre-Antoine Champin

RDF in one slide

An abstract data model
Information is broken down into statements (a.k.a. triples)...
... of the form subject predicate object
A set of triples can be represented as a graph

An example RDF graph showing a person named 'Pierre-Antoine Champin' working for an organization named 'World Wide Web Consortimu'

JSON-LD in the RDF ecosystem

This figure shows the different parts of the eco-system: RDF abstract syntax is surrounded by RDF semantics and RDF abstract syntaxes. SPARQL, SHACL and others are linked to the abstract syntax. RDF-S, OWL and others are linked to the semantics.

DDI-CDI to JSON-LD?
(short answer)

First map DDI-CDI to RDF (using OWL, SHACL...)
Then you can represent it in JSON-LD (or Turtle, or RDF/XML...)
Problem solved! 🎉

But there is more to it...

Straightforward JSON-LD serialization

[
  {
    "@id": "https://json-ld.org/playground/#pa",
    "@type": [ "http://schema.org/Person" ],
    "http://schema.org/name": [
      { "@value": "Pierre-Antoine Champin" }
    ],
    "http://schema.org/worksFor": [
      { "@id": "https://json-ld.org/playground/#w3" }
    ]
  },
  {
    "@id": "https://json-ld.org/playground/#w3",
    "@type": [ "http://schema.org/Organization" ],
    "http://schema.org/name": [
      { "@value": "World Wide Web Consortium" }
    ]
  }
]

A more idiomatic JSON representation

{
  "id": "#pa",
  "type": "Person",
  "name": "Pierre-Antoine Champin",
  "worksFor": {
    "id": "#w3",
    "type": "Organization",
    "name": "World Wide Web Consortium"
  }
}

The four main algorithms of JSON-LD

Benefit of the compact form

A compact JSON-LD document can be used from different points of view: either as representing an RDF graph, or as representing 'plain' JSON data

Keeping track of the context

The JSON-LD context can be

Embedded in the data:
"@context": {...},
Referenced by the data:
"@context": "http://example.org/my-context.jsonld",
Provided in an HTTP Link header
Provided off-band

What's in a context (1)

Mapping from short names to IRIs (or JSON-LD keywords)

{
  "@context": {
    "id": "@id",
    "type": "@type",
    "Person": "http://schema.org/Person",
    "name": "http://schema.org/name",
    "worksFor": "http://schema.org/worksFor",
    "Organization": "http://schema.org/Organization"
  },
  "id": "#pa",
  "type": "Person",
  "name": "Pierre-Antoine Champin",
  "worksFor": {
    "id": "#w3",
    "type": "Organization",
    "name": "World Wide Web Consortium"
} }

What's in a context (2)

Definition of prefixes to shorten IRIs
Disambiguation of the meaning of attribute values
- IRI
- literal with given datatype or language tag
- set or list of values (JSON arrays)
Support for idomatic structures
and more...

See

What's not in a context

JSON-LD is not a general purpose transformation language.

It is opinionated
It favours simplicity over expressivity
It focuses on real uses cases and patterns

DDI-CDI to JSON-LD?
(longer answer)

Map DDI-CDI to RDF (using OWL, SHACL...)
Design an idiomatic JSON format for DDI-CDI
Then build a JSON-LD context to bridge between the two
(might require some adjustments)

Going further: RDF-star

A community effort to extend RDF

with the ability to make statements about statements.

See

JSON-LD-star

{
  "id": "#pa",
  "type": "Person",
  "name": "Pierre-Antoine Champin",
  "worksFor": {
    "@annotation": {
      "startDate": "2021-02-21"
    },
    "id": "#w3",
    "type": "Organization",
    "name": "World Wide Web Consortium"
  }
}

See

Any question?

Appendix

Framing

Actually, to go from the ugly JSON-LD serialization to the idiomatic one in the introductory example, the compaction algorithm is not sufficient.

The JSON-LD specification defines another algorithm called framing, which is complementary to compaction.

In a nutshell, compaction works at the level of individual attributes, while framing can reorganize the structure of the whole document.

See

What's in a context (example)

Desambiguation of the values (node type, datatype, langue)

An example RDF graph showing the #w3c was founded by #tbl on 1994-10-01, and that its slogan is 'Leading the web to its full potential'

NB: also notice that the "short names" do not need to be similar to the IRI.

{
  "@context": {
    "id": "@id",
    "s": "http://schema.org/",
    "xsd": "http://www.w3.org/2001/XMLSchema#",
    "founder": { "@id": "s:founder",
                 "@type": "@id" },
    "founded": { "@id": "s:foundingDate",
                 "@type": "xsd:date" },
    "motto": { "@id": "s:slogan",
                 "@language": "en" }
    },
  "id": "#w3c",
  "founder": "#tbl",
  "founded": "1994-10-01",
  "motto": "Leading the web to its full potential"
}