Re: Semantics TF discussion

> On 11. Dec 2024, at 10:42, Franconi Enrico <franconi@inf.unibz.it> wrote:
> 
> On 10 Dec 2024, at 23:15, Thomas Lörtsch <tl@rat.io> wrote:
>>> IMHO, we shouldn’t have such an exception, so to capture the (debatable) case:
>>> 
>>> :s rdf:reifies “John loves Mary” .
>> 
>> I don’t understand what you are arguing for. In my understanding nothing but a triple term makes sense in the range of rdf:reifies. 
> 
> There are two issues here:
> do we allow literals to be object of rdf:reifies?
> IMHO, this should be possible in a “liberal” approach. Otherwise we go straight to some less liberal well-formedness - which is also fine.

RDF allows all kinds of nonsense, like 

    _:x rdfs:subPropertyOf "__ᓚᘏᗢ__"^^xsd:string .

so why should we bother to disallow nonsensical uses of rdf:reifies? We need to define and properly explain what a triple term is and what its use with rdf:reifies means, but I don’t see why we would need to explain what else a triple term is not and what other uses of rdf:reifies are meaningless.

However, maybe you see sensible uses of rdf:reifies besides reifying triple terms?

> should any object of rdf:reifies be a rdf:Proposition (or some other special name)?
> I’d say yes, but of course if you expect that triple terms should appear only as object of rdf:reifies (which we should not!), then the notion of rdf:Proposition would be redundant in this case since it coincides with triple term. 

IMO the object of an rdf:reifies relation is explained well enough if the rdf:reifies relation is explained well enough. I don’t think it needs a special type.

My understanding was that the proposal is to rename rdf:TripleTerm to rdf:Proposition, not to have both. I see some advantages to rdf:Proposition
- it nicely contrasts with rdf:Statement (still needing an explanation, of course)
- it emphasizes the semantics aspect over the syntactic aspect (which I like)
But emphasizing the syntactic aspect makes sense too, especially since the other terms with which it is grouped - IRI, blank node and literal - all emphasize the syntactic aspect. So maybe it’s better to go with rdf:TripleTerm. It would of course be great to be able to call it rdf:Triple, but IMO that creates much too much potential for confusion, especially since triple terms are not asserted.

> A triple term may play several roles. 
> When object of rdf:reifies, it plays the role of being reified. 
> In the future, we can imagine that a triple term plays (also) the role of an element of a named graph - and we should not block this possible future.
> 
>  For this reason, I believe that a triple term should be marked accordingly: 
> by using rdf:Proposition (if the name seems misleading, we can choose another one) if it is object of rdf:reifies; 
> in a possible future by using, e.g., rdf:TripleInNamedGraph, when used for this purpose.
> 
> Example:
> 
> _:r rdf:reifies <(:john :wife :mary)>.
> _:r rdf:reifies <(:mary :husband :john)>.
> _:r a :marriage.
> <(:john :wife :mary)> a rdf:Proposition.
> <(:mary :husband :john)> a rdf:Proposition.
> 
> (in a hypothetical future)
> _:g rdf:hasTriple <(:john :wife :mary)>.
> _:g rdf:hasTriple <(:mary :husband :john)>.
> _:g a rdf:NamedGraph.
> <(:john :wife :mary)> a rdf: TripleInNamedGraph.
> <(:mary :husband :john)> a rdf: TripleInNamedGraph.

I’m totally open to such future uses of triple terms, but IMO it makes more sense to type the resulting constructs that are derived from triple terms, like refications, graphs, etc, than the triple terms themselves. I.e. no matter if one derives a reification, a graph or whatever else, it is always derived from a triple term. That triple term always has the same type.

Adding one statement to your example:

    :John a rdf:IRIusedInObjectPositionOfATripleTerm .

In other words: who does such a thing ;-) 

Some axiomatic triples won’t hurt, but they should type the result, not the building block used to construct that result - e.g.

    rdf:reifies rdfs:domain rdf:Reification ;
                rdfs:range rdf:TripleTerm .
    rdfs:states rdfs:domain rdf:StatedReification ;
                rdfs:range rdf:TripleTerm .
    rdf:contains rdfs:domain rdf:NamedGraphTerm ;
                rdfs:range rdf:TripleTerm . 


>>> Also, note that not all triple terms are also propositions: this happens only if a triple term appears as an object of rdf:reifies.
>> 
>> Really? In my understanding a triple term always, in any position and with any predicate, 
>>     describes a statement (of a certain type) without stating it. 
> 
> Sure, but here rdf:Proposition was meant to mark triple terms used by some reification process. Maybe we should change the name.

Or just omit it, as argued above.

>> And in exactly that way it is used in the range of rdf:reifies. So why give it another type there?
>> I understood our discussion in last Friday’s Semantics TF that all triple terms are of type rdf:Proposition, to dismabiguate them from rdf:Statement. 
> 
> Sure, the idea was that triple terms used in a reification are rdf:Proposition, exactly to disambiguate them from rdf:Statement, but this does not imply that all triple terms, regardless from their use are rdf:Proposition.

Also already argued above.

Bes.t


> cheers
> —e.
> 
>>>> On 09/12/2024 17:23, Franconi Enrico wrote:
>>>>> We had an interesting Semantics TF meeting last Friday, and we concluded a couple of things:
>>>>> 1. It would be interesting to consider, instead of the currently
>>>>>    approved alternative baseline, a simpler more “liberal” approach: a
>>>>>    new property rdf:reifies is introduced, which is used to express
>>>>>    reification - namely the relationship between a reifier (sometimes
>>>>>    called “token") and an abstract triple term. In order to express
>>>>>    reification, *only* the rdf:reifies (or any sub property of it)
>>>>>    could be used. Other usages of triple terms (namely without being
>>>>>    object of a rdf:reifies property or of any of its subproperty) is
>>>>>    not forbidden (hence the name: “liberal” approach), but they would
>>>>>    not be having the meaning of a reification.
>>>>> 2. We discussed the possible type of the object of rdf:reifies: it
>>>>>    could be rdf:Proposition.
>>>>> This is captured - with many details still to be discussed - in <https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22liberal-baseline%22 <https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22liberal-baseline%22>>.
>>>>> —e.
>>>> 
>>>> 
>>> 
>> 
>