Leaving presentation mode.
RDF-star Semantics Task Force
Enrico Franconi
Andy Seaborne, Antoine Zimmermann, Dörthe Arndt, Gregg Kellogg, Niklas Lindström, Peter Patel-Schneider, Pierre-Antoine Champin, Ted Thibodeau
★ W3C TPAC 2023, Seville, 11–15 September ★
Contents
- Scope of the Task Force
- Community Group Cases
- Some additional Cases
- A catalogue of Behaviours
- Current Proposals
- Conclusions
Scope of the Task Force
- Extend the model-theoretic semantics of RDF 1.1 to support quoted triples.
- Define a conservative extension - i.e., maintaining compatibility also at the level of models and therefore entailment.
- Provide a provably correct entailment-preserving encoding of a RDF-star graph in a RDF 1.1 graph.
- Take into account existing implementations.
- Take into account the outcome of the Community Group.
Community Group Cases
Opaque IRIs and Literals, transparent BNodes, and opaque properties:
<< dbr:Linköping dbo:populationTotal "104232"^^xsd::nonNegativeInteger >>
:type rdf-star:triple ;
:source <https://dbpedia.org/data/Linköping.ttl> .
Transparent IRIs, Literals, and BNodes, and opaque properties:
:measuredOn rdf:type rdf-star:TransparencyEnablingProperty .
<< dbr:Linköping dbo:populationTotal "104232"^^xsd::nonNegativeInteger >>
:type rdf-star:statement ;
:measuredOn "2010-12-31"^^xsd:date .
Some additional Cases
Transparent IRIs, Literals, and BNodes, and transparent properties:
<< :john :teaches :cs101 >>
rdf:type rdf-star:event ;
rdf:type :teaching ;
dct:Location dbr:Stanford_University .
Locally opaque IRIs, Literals, and BNodes, and locally opaque properties:
:paul :believes << :ted owl:same-as :theodore >>
:paul :believes << :mary :loves :ted >>
:john :believes << :mary :loves :ted >>
⊨ :paul :believes << :mary :loves :theodore >>
⊭ :john :believes << :mary :loves :theodore >>
...
A catalogue of Behaviours
IRIs, literals, BNodes, and properties in quoted triples can be:
- transparent,
- opaque,
- locally opaque,
- any combination of the above.
A use cases should identify which combination of the above supports.
IRIs, literals, BNodes, and properties in quoted triples can be:
A detailed comparison is being initiated now.
Conclusions
- We have a good understanding of the different behaviours,
- we can categorise use cases,
- we are comparing different approaches to model-theoretic semantics.
- Some open questions:
- Decision (by the WG) on which behaviours and syntaxes to standardise.
- Connection with N3 semantics.
- 🐘 Extension to quoted (named) graphs?