W3C

– DRAFT –
RDF-star SemanticsTF

24 January 2025

Attendees

Present
AndyS, AZ, doerthe, Dominik_T, eBremer, enrico, gkellogg, gtw, james, ktk, niklasl, ora, pchampin, pfps, Souri, TallTed, tl
Regrets
-
Chair
-
Scribe
AndyS

Meeting minutes

Discuss on the open issues and PR related to semantics in the RDF-semantics document

<doerthe> w3c/rdf-semantics#69

<gb> Issue 69 modification to new semantics to permit entailment that every triple term is a proposition (by pfps)

Oh.

<enrico> w3c/rdf-semantics#69

<enrico> w3c/rdf-semantics#71

<gb> Issue 71 Fixing the definition of instance of a graph to consider triple terms (by franconi) [spec:substantive]

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

<Zakim> AndyS, you wanted to ask if this definition of triple term is written done (in public) yet

<enrico> https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22liberal-baseline%22#rdfs-semantics

w3c/rdf-semantics#49

<gb> Issue 49 Define an interpretation of Triple Terms (by niklasl) [needs discussion]

The "RDFS entailment for SPARQL" should be defined to have finite answers

<doerthe> but we are defining a calculus here

(a requirement of all entailment regimes)

<doerthe> it should be complete

<niklasl> rdf:_1 a rdf:Property # ...

<doerthe> question is rather is "s p o" entailed

quite - this is not the SPARQL doc on enailment regimes. this is RDF processors being accessed by SPARQL not SPARQL processors.

<enrico> yes, AndyS

<niklasl> Right, rdf:type is in RDF entailment.

<niklasl> IMHO, a triple term where the predicate does not denote a property is not a proposition; it does not exist -- i.e.it is nonsensical.

<enrico> w3c/rdf-semantics#68

<gb> Pull Request 68 Add simple semantics to RDF-semantics (by franconi) [propose closing] [spec:substantive]

<enrico> to niklasl 😱

Are there any general purposed & extant SPARQL engine implementing entailment regimes? (for getting through W3C process)

<Zakim> pchampin, you wanted to point out that, if the predicate of triple terms are required to be properties, then <<( s p o )>> a rdfs:Proposition can not be inferred for arbitrary s,p,o (from the empty graph)

<enrico> w3c/rdf-semantics#69 (comment)

<gb> Issue 69 modification to new semantics to permit entailment that every triple term is a proposition (by pfps)

<enrico> We agree with the proposal at w3c/rdf-semantics#69 (comment)

<enrico> +1

<niklasl> +1

<doerthe> +1

enrico: RE is total.

<pfps> +1

enrico: implicit - always says "partial" if it is not total

<enrico> w3c/rdf-semantics#69 (comment)

<enrico> https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22liberal-baseline%22#rdfs-semantics

https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22liberal-baseline%22/d8f94f1bf4551ea7f5135f37f468a64cef0e684d

<niklasl> the blank node trick?

<TallTed> I would prefer that nothing be implicit; that we say explicitly whatever we want others to infer

https://www.w3.org/TR/sparql12-entailment/

<pfps> https://www.w3.org/TR/sparql12-entailment/#RIFFiniteAnswers

https://www.w3.org/TR/sparql12-entailment/#RDFSEntRegime

<Souri> Would the correct answer from SPARQL be finite for { ?s rdf:type rdfs:Resource } given an empty RDF graph?

Souri - SPARQL is based simple entailment together with a matching step - here, empty graph means results in zero rows.

<enrico> From https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22liberal-baseline%22#rdfs-semantics: we accept rdfs14, and we will have in appendix A its approximation in the strict syntax

<niklasl> +1

<tl> +1

Link to rdfs14: https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22liberal-baseline%22#rdfs-semantics::~:text=S%20RDFS%20entails-,rdfs14,-xxx%20rdf%3Atype

<doerthe> +1

<enrico> complete entailment patterns for RDFS:

<enrico> rdfs4a-ts, rdfs4b-ts, rdfs14

<enrico> Approximated entailment patterns in appendix A:

<enrico> if the triple structure appears in S

<enrico> sss aaa <<(xxx yyy zzz)>>

<enrico> then S RDFS entails

<enrico> sss aaa _:b.

<enrico> _:b rdf:type rdfs:Proposition.

<enrico> we do not make any claim about completeness

<enrico> at this point. Hopefully somebody will prove it

<pfps> w3c/sparql-entailment#37

<gb> Issue 37 inifinite results in entailment regimes (by pfps) [spec:editorial]

<doerthe> If you don't give a sh*t about metamodelling, you don't care about the domain of RE being IRxIRxIR ;) because the IP is only used for metamodelling

<tl> i'm all for nesting - i.e. i find it really useful at least 2 levels deep -, and since implementations seem to have had no problem implementing it for RDF* i don't see why we should think about dropping this feature

<niklasl> w3c/rdf-semantics#71

<gb> Issue 71 Fixing the definition of instance of a graph to consider triple terms (by franconi) [spec:substantive]

w3c/rdf-concepts#144

<gb> Issue 144 the term identity comparison must be defined for triple terms (by lisp)

As PFPS says - no choice points/backtracking.

<gb> Issue 150 Nature / role of triples (by pchampin)

As RDF terms bnode are concrete. Semantics may wish to write entailment equivalent.

w3c/rdf-concepts#150

<niklasl> The semantics now contain "the interpretation of triple terms" and "ground triple term"; may need to be looked at?

<niklasl> ... *if* 150 is decided.

Minutes manually created (not a transcript), formatted by scribe.perl version 242 (Fri Dec 20 18:32:17 2024 UTC).

Diagnostics

Succeeded: s/done/down/

Succeeded: s/shit/sh*t/

Succeeded: s/dan't/don't/

Succeeded: s/Semantcis/Semantics/

Succeeded: s/contain "he interpretation/contain "the interpretation/

No scribenick or scribe found. Guessed: AndyS

All speakers: enrico

Active on IRC: AndyS, doerthe, enrico, niklasl, pchampin, pfps, Souri, TallTed, tl