Re: Decision from the Semantics TF: liberal baseline

> On 9. Jan 2025, at 15:31, Peter F. Patel-Schneider <pfpschneider@gmail.com> wrote:

> I think that the discussion is best formed around which entailements we want, and that the entailments in question all are related to whether the mere presence of triple terms in a graph creates entailments as if they were asserted triples in the graph.
> 
> Here are some of the potential entailments:
> 
> :a rdf:reifies <<( :b :c :d )>>
> simple entails
> :b :c :d .
> I say no to this one.

Agreed. I guess that’s uncontroversial.

> :a rdf:reifies <<( :b :c :d )>>
> RDF entails
> :c rdf:type rdf:Property .
> I say no to this one as well.

Agreed: let triple terms be entities that are closed to introspection.

> :a rdf:reifies <<( :b :c :d )>>
> RDFS entails
> :b rdf:type rdfs:Resource .
> I say yes to this one, because everything is a resource so having an IRI in a triple term doesn't really add anything.

That too me stands in contradiction to the one before. Isn’t everything in predicate position an rdf:Property (or illegal)? Also, what’s the point in making such brittle distinctions when the result is not of any interest.

> :a rdf:reifies <<( :b rdf:reifies :d )>>
> RDFS entails
> :d rdf:type rdfs:Proposition .
> I say no to this one.

Agreed, as again: let triple terms be entities that are closed to introspection.

> :a rdf:reifies <<( :b :c "hi"^^xsd:string )>> .
> extended-RDFS entails
> "hi"^^xsd:string rdf:type xsd:string .
> I say yes to this one, because strings are always instances of xsd:string in RDFS so having a string in a triple term doesn't really add anything.

I see again a contradiction, same argument as above.

> :a rdf:reifies <<( :b rdf:reifies <<( :e :f :g )>> )>> .
> RDFS entails
> <<( :e :f :g )>> rdf:type rdfs:Proposition .
> I say yes to this one, again because triple terms are always instances of rdfs:Proposition in RDFS so having a triple terms inside a triple term doesn't really add anything.

I see again a contradiction, same argument as above.

I think it makes more sense to catgorically say that triple terms are off limits to introspection. They can be converted to regular triples via the unstar mapping and then, on the level of regular triples be described, reasoned upon etc.

Thomas
> 
> 
> 
> 
> 
> peter
> 
> 
> On 1/9/25 8:56 AM, Doerthe Arndt wrote:
>> Hi all,
>> I am not sure to which message I should answer, so I try to go through the different points:
>> - *"**if the triple structure appears in S“*: I fully agree with Peter that "if the triple structure appears in S“  can not be put into the spec without further formalization, I still hope to get completely rid of it. If it keeps being vague, there is no point in having entailment patterns in the first place.
>> -***better way to get the effect of rdfs:Proposition: *I am not sure whether I fully understand Peter, but I would read this as defining the mapping RE: IR x IP x IR -> TP, where TP is our set of propositions (subset of IR), then on the model theory, we would say:
>> x is in TP if and only if <x, I(|rdf:Proposition|)> is in IEXT(I(|rdf:type|))
>> Peter, was that what you meant? (The yellow comes from copy-paste, please ignore)
>> - *super range of rdf:reifies: *I think I already expressed that I am against that as well,
>> - *metamodelling* as (quote below)
>>> Just to open a novel can of worms:
>>> there is a bunch of metamodelling semantic conditions in RDFS which currently do apply only at top level, and we have to decide whether to generalize to triple terms at arbitrary nesting.
>>> 
>>> For example:
>>> 
>>> :a :b <<(:c rdf:type :d)>>.
>>> 
>>>    should or should not RDFS-entail 
>>> :d rdf:type rdfs:Class.
>>> 
>>> :a :b <<(:c rdfs:subclass :d)>>.
>>> 
>>>    should or should not RDFS-entail 
>>> :c rdf:type rdfs:Class.
>>> :d rdf:type rdfs:Class.
>> I am strongly against that. What if we have, for example:
>> :x rdf:reifies <<(:c rdfs:subclass :d)>>;
>> a :Lie.
>> I could come up with stronger examples, but I am also against that.
>> - *Literals: *we should also modify rdfD1
>> */- "Importance of /**metamodelling apparatus discussed“: * we need to have the entailment patterns (these have been in the spec and should remain there) and these need to be correct, so we have to discuss them at some point. In my opinion, there are also very helpful to see what the semantic definitions mean.
>> Kind regards,
>> Dörthe
> 
>

Received on Friday, 10 January 2025 14:48:02 UTC