RDF-star / RDF 1.2

RDF timeline

Statements about statements

Up until now done by “modeling” RDF statements in RDF

Reification: cumbersome, misunderstood, unpopular

• one of the most persistent criticisms of RDF, particularly when comparing to Labeled Property Graphs (LPGs)
• could be seen as the RDF counterpart of Labeled Property Graph “edge properties” (RDF is more expressive, though)

Over the years, many ideas and proposals on how to fix this

• most notably, “Reification Done Right” by Hartig & Thompson in 2014
• RDR eventually parlayed into “RDF-Star”, first a CG, now a WG

Some use cases

Provenance

Qualifying statements

“Marginalia”

Modalities (beliefs, etc.)

Also: bringing RDF closer to LPGs (aka “alignment”)

Challenges

Triples in an RDF graph are unique (remember: set semantics)

:ElizabethTaylor :married :RichardBurton

• we want to qualify this with a year
• but... Taylor married Burton twice (1964 and 1975)

Breaking existing RDF (1.1) semantics is highly undesirable

Introducing new semantics could have consequences for RDFS & OWL

Relationship between statements and named graphs?

• is a reified statement a singleton named graph?

Semantics proposal

We have extended the RDF 1.1 semantics with “triple terms”

• triples are formed using IRIs, blank nodes, literals, and triple terms
• an occurrence of a triple is identified by a reifying resource (aka a “reifier”) related to a triple term

We have agreed on a minimal baseline as a means of comparing proposals

• we introduce minimal semantics and semantics for the well-formed fragment of RDF (we call this “RDF semantics”)

We have two alternate, extended baselines we are still discussing

• these essentially differ in how a reifier is related to triple terms

Syntax proposal

We extend Turtle with some “syntactic sugar”:

Occurrence syntax (or "unasserted syntax", still debating this):

<< :s :p :o ~ :id >> :pp :oo

Annotation syntax:

:s :p :o {| :pp :oo |}

“Triple term” syntax (only for internal use):

<<(:s :p :o)>>

(Niklas Lindström will get into the details in the next presentation.)

Current status

It took us a long time to agree on the “big picture” (but we now have more or less reached consensus within the WG)

RDF 1.1 semantics has been extended (this was the really important part)

We have lots of documents that now need to all be updated (essentially all RDF and SPARQL specs)

“RDF-Star” name will go away and we get “RDF 1.2” and “SPARQL 1.2”