Reifiers & Graphs

RDF-star WG

TPAC 2024
Anaheim CA, USA
hybrid meeting
23–27 SEPTEMBER 2024

(Slides by Niklas Lindström, National Library of Sweden.)

Preceded by: Reifiers

Is a Reifier a Graph?

Not per se.

The triple terms reified are transparent; their constituent subject, predicate and object terms denote entities within the interpretation.

The reifying purchase is in the same domain of discourse as Alice, the plush lion, the store, and their relationships.

Triples In The Graph

A reifier is not necessarily defined by the triple or triples it reifies. So while a set of the triples reified can be seen as some graph, this does not identify the reifier.

For example, a purchase is a conceivable concretization of a bought and sold relationship, but it is not intrinsically defined by them. It has other properties, such as a point in time, a location, etc.

Unlike Graphs

A graph is a set of triples. A graph of one triple is a different entity than one of two, even if the former is a subset of the latter.
Reifiers have an rdf:reifies relation to each of its triples. (Just as Alice has two parents, not a set of parents.) Each relationship holds individually, monotonically.

Using OWL, this can be constrained, e.g. on cardinality:

ex:StatementToken rdfs:subClassOf [ a owl:Restriction ;
      owl:onProperty rdf:reifies ;
      owl:cardinality 1 ] .

RDF Named Graphs

A named graph is a pair of a name (IRI or bnode) and a graph (set of triples).
There is no normatively defined relation between a named graph and the resource denoted by this name.

GRAPH <document> {
  <document> a :Document .
}

An Insulating Gap

While the name of a named graph may denote anything, including a reifier, there is no relation from it to the graph, nor to the triples.

<purchase1> a :Purchase ;
  :date "2024-06" .

GRAPH <purchase1> {
  :Alice :bought :LennyTheLion .
  :ToyStore :sold :LennyTheLion .
}

This prevents further modelling of the relationship between the reifier and its reified triples (such as cardinality, ranges or property chain axioms).

Outside of Graphs

Named graphs are for management of graphs in datasets, external to their interpretation(s).

There is no normatively defined interpretation for datasets.

Can be used in practice as a union (see "union default graph" modes of many SPARQL stores, for querying over the union of all graphs in a dataset).
Or as separate, isolated interpretations (e.g. Verifiable Credentials).

(See also the RDF 1.1: On Semantics of RDF Datasets note.)

Reifiers Unify Usage Patterns

The domain of rdf:reifies is open (of any type). Examples include:

A token in a source document (classic reification rdf:Statement).
An utterance or linguistic expression encoded as multiple related triples ("a person named Alice Liddell").
A source message or document ("T indicates that A bought B from C").
A circumstance or event (a Collaboration, a Purchase, etc.).

This unifies the earlier dichotomy between classic reification and "that which named graph names may denote".

A Note About Abstract Graphs

The graph itself—as a mathematical set of triples—is reasonably not a concretization of itself.

It is a set of abstract triples.

Triple terms may play a role in defining such. Cf. Notation 3.

Thank You!

Time for discussion?