Re: transparency and entailment

> On 19. Feb 2024, at 12:52, Franconi Enrico <franconi@inf.unibz.it> wrote:
> 
> 
> 
>> On 17 Feb 2024, at 20:18, Peter F. Patel-Schneider <pfpschneider@gmail.com> wrote:
>> 
>> I think that this characterization is not sufficient for transparency. Consider the CG semantics, which is a macro-expansion that then uses the usual RDF semantics, which does satisfy your criterion.  But the CG version of quoted triples is not transparent.
> 
> My characterisation is sufficient whenever RDF has a direct model-theoretic semantics, which the CG semantics is not (it is based on a translation).
> RDF-star will have a direct model theoretic semantics, if I am going to remain in the WG :-)
> —e.

Your characterization (below) doesn’t capture my intuition of referential transparency, whereas Peter’s (further below) does. More concretely your characterization doesn’t seem to capture the notion of co-reference: a syntactic representation of a referent can be exchanged with another syntactic representation of a referent iff they both refer to the same entity in the realm of interpretation. I don’t assume that it's impossible to represent such a notion in a model-theoretic semantics (rather it seems to me that formalizing the opposite - referential opacity - is more difficult within the framework of the RDF model theory). 

Thomas


>> peter
>> 
>> PS:  I suspect that you would want to include literals as well.
>> 
>> On 2/17/24 10:12, Franconi Enrico wrote:
>>> To me, transparency means:
>>> given a graph G, II is the set of all IRIs appearing in G and BB is the set of all bnode symbols appearing in G.
>>> Then, ∀ i∈II and b∈BB, i and b have the same denotation non matter where they appear within the graph.
>>> I guess that your definition below is somehow different, but probably it boils down to mine, which is more clear, I guess.
>>> —e.
>>>> On 16 Feb 2024, at 18:04, Pierre-Antoine Champin <pierre-antoine@w3.org> wrote:
>>>> 
>>>> Peter,
>>>> 
>>>> On 09/02/2024 20:24, Peter F. Patel-Schneider wrote:
>>>>> There was some discussion of transparency in the semantics call today, with disagreement over just what transparency means.
>>>>> 
>>>>> My view is that transparency (for well-formed graphs) means that entailments are exactly the same if a subject, predicate, or object in a quoted triple is replaced by a semantically identical identifier.  So if an option for << e | s p o >> is transparent in D-entailment then
>>>>> 
>>>>> << :e | :s :p "4"^^xsd:integer >> :a :b .
>>>>> 
>>>>> entails
>>>>> 
>>>>> << :e | :s :p "04"^^xsd:integer >> :a :b .
>>>>> 
>>>>> in that option.
>>>> 
>>>> that's also my interpretation of "transparency".
>>>> 
>>>> (and I assume that the entailment in your example above works both ways)
>>>> 
>>>>> 
>>>>> 
>>>>> peter
>>>>> 
>>>> <OpenPGP_0x9D1EDAEEEF98D438.asc>
>

Received on Monday, 19 February 2024 12:08:40 UTC