– DRAFT –
RDF-star SemanticsTF

discussion on metamodeling semantics/entailments for the liberal baseline

https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22liberal-baseline%22

enrico: IR x IP x IR => IR x (IR U IP) x IR

pfps: Everything in simple semanticsneeds to be (IR U IP)

enrico: IR x IP x IR => (IR U IP) x (IR U IP) x (IR U IP)

enrico: tripleStructure is a 3-tuple of RDF terms.

<niklasl> +1 re. name of abstract syntax

AndyS: there is a problem with the name "abstract syntax" in the wiki page about liberal semantics
… this is different from the abstract syntax defined in rdf-concepts, and the relation should be clarified

<niklasl> Just use symmetric RDF (amended with bnodes allowed in predicate position).

<pchampin> enrico: the abstract syntax in the wiki page defines things that are not in RDF Concepts, which we need to define the semantics

<pchampin> ... for me the reference is the wiki page, the RDF Concept document should be aligned afterwards

<niklasl> The abstract syntax is for RDF graphs. RDF graph data model?

<pchampin> AndyS: RDF Concepts defines a "data model" more than an "abstract syntax", but that ship has sailed

<AndyS> scribe:

<pfps> it appears to me that the "abstract syntax" in the liberal baseline is the same as "RDF Graphs" in Concepts

pchampin: wiki "abstract syntax" is reformulation of RDF Concepts data model/abstract syntax
… modulo triple terms in the subject position and renaming, they align.

enrico: can't vote next week on the concept document
… need time to work on the concepts document

gkellogg: Don't see we have agreed on subject/triple terms yet
… maybe symmetric for the liberal baseline.
… if the RDF data model has literals/triple terms in the subject position then users expect this in the concrete syntaxes.

<pchampin> I like the idea of defining the semantics on the symetric abstract syntax, but that would need to the symetric abstract syntax to be normative... to we want that?

Possibility -- make it normative in RDF semantics. (also - the bnodes as predicates corner case)

pchampin: c.f. social meaning of IRIs and formal semantics for IRIs

<niklasl> I am more bothered by the disconnect.

Maybe -- wiki "abstract syntax" -> "semantics data model" or "semantics abstract syntax " or ?

<niklasl> in the "real world", not the *real* world. The model.

<niklasl> (better: the domain of discourse)

<niklasl> Too weak.

simple semantics

<niklasl> You can build multiple worlds within one world. It just requires more blank nodes.

<niklasl> The social contract is imperative for interoperability.

<tl> @niklasl e=mc^2 is a proposition, and may also stand in for a whole theory

<niklasl> "abstract graph syntax" and "abstract model syntax" ?

<niklasl> +1 for the "KR stuff" being in introduction in semantics; perhaps concepts could reference it though?

<pfps> Ooops, the syntax in the liberal baseline does allow triple terms in subject position.

<enrico> ofc, pfps

<Zakim> pfps, you wanted to ask about what we will propose to vote on next week

<niklasl> +100

<niklasl> Yes, in a universe where properties are not resources, no triple can describe them.

<niklasl> Making a statement is putting a triple into a graph. That encodes the meaning that the proposition is true in the domain of discourse (described by the graph).

<doerthe> then I can't assert it, but I don't see the problem

<niklasl> RDF semantic condition one: x is in IP if and only if <x, I(rdf:Property)> is in IEXT(I(rdf:type))

<pchampin> x is in IP if and only if <x, I(rdf:Property)> is in IEXT(I(rdf:type))

<pchampin> since IEXT \subseteq IRxIR, therefore x \in IR

<niklasl> Further emphasized in RDFS of course rdf:Property rdfs:subClassOf rdfs:Resource . :)

<niklasl> :r rdf:reifies <<( rdfs:Resource rdfs:Resource rdfs:Resource )>> . # :/

<pfps> There is nothing in the normative parts of RDF Concepts that requires that the second element of a triple is a property.

<pfps> From Concepts: "Asserting an RDF triple says that some relationship, indicated by the predicate, holds between the resources denoted by the subject and object. This statement corresponding to an RDF triple is known as an RDF statement. The predicate itself is an IRI and denotes a property, that is, a resource that can be thought of as a binary relation."

<pfps> So it is only when a triple is asserted that its second element is a property.

<doerthe> for the record, I only lean towards Peter's position, but I am fine with both solutions

Concepts quote: <![[

An RDF triple (usually called "triple") is a 3-tuple (s, p, o) with the following characteristics:

* s is an IRI or a blank node.

* p is an IRI.

* o is an IRI, a blank node, a literal, or a triple term.

]]> and RDF 1.1 "An RDF triple consists of three components:"

<niklasl> A syntactic restriction could even drop the triple containing the "broken" triple term?

<niklasl> I'm prepared to pay that price.

<niklasl> Otherwise we should just add LISP to RDF.

<tl> +1 to adding LISP to RDF

<niklasl> We're even doing it, but a weird three-tuple version of it, rather than *pairs*. Add those, and enjoy the ensuing computational complexity.

<enrico> https://w3c.github.io/rdf-concepts/spec/#section-triple-terms-reification

<pchampin> https://www.w3.org/TR/rdf12-concepts/#resources-and-statements, 2nd paragraph

<niklasl> How could you have a triple just lying around in RDF 1.1?

<doerthe> it is not about use cases it is about what we have to entail

<pfps> Triples just *do* lie around. They don't have to be in the RDF graph under consideration. Only triples in the graph get this treatment.

<niklasl> How are the other triples treated?

<pchampin> the use case is talking about triples asserted by someone else, but that someone else is talking (syntactically valid) nonsense

<doerthe> It boils down to one question in my opinion:

<doerthe> :a rdf:reifies <<( :b :c :d )>>

<doerthe> RDF entails

<doerthe> :c rdf:type rdf:Property .

<doerthe> (taken from Peter's mail)

<doerthe> So, what do we want there?

pfps: suggestion -- wiki page upto (and excluding) "RDF semantics" + note wiki 'abstract syntax' is more liberal than concepts and has no effect

<pchampin> PROPOSAL: apply liberal baseline to rdf-semantics up to "RDF-Semantics", referring to rdf-concepts for abstract syntax

<pfps> +1 for this proposal

https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22liberal-baseline%22/3d9339abfd31edcbd0a684f8b9930df7176b80b9

s!https: //github.com/w3c/rdf-star-wg/wiki/RDF-star-%22liberal-baseline%22/3d9339abfd31edcbd0a684f8b9930df7176b80b9!!

<pfps> +1

<pchampin> PROPOSAL: apply liberal baseline to rdf-semantics up to "RDF-Semantics", referring to rdf-concepts for abstract syntax

<pfps> +1

<pchampin> +1

<niklasl> +1

<tl> +1

<doerthe> +1

<gkellogg> +1

<TallTed> +1

<AndyS> +0

<enrico> +1

<niklasl> Yes +1 for succinctness.

RESOLUTION: apply liberal baseline to rdf-semantics up to "RDF-Semantics", referring to rdf-concepts for abstract syntax

AndyS: unsure about details of today's conversations but ATM looks good.