Re: Consolidating triple/edges -- named occurrence version

> On 5. Jan 2024, at 14:41, Franconi Enrico <franconi@inf.unibz.it> wrote:
> 
>> On 5 Jan 2024, at 13:30, Thomas Lörtsch <tl@rat.io> wrote:
>> 
>> Is the following 
>>>  << :wed-1 | :liz :spouse :richard >> .
>> an "occurrence term" (my label), and as such is asserted?
> 
> Nope. A “term” is never asserted, it is just an expression denoting a resource.

Aha. I assumed it’s the syntactic equivalent to the shortand syntax, just with chevrons.

>> so different from an unasserted triple term 
>>    << :liz :spouse :richard >>
>> and syntactic sugar for 
>>    :wed-1 rdfx:occurrenceOf << :liz :spouse :richard >>
>> ?
> 
> In an approach where triple terms are part of the language as terms, they again do denote resources and they are not asserted.

Sorry for the insistency: I do now assume that

    << :wed-1 | :liz :spouse :richard >> .

(with the dot at the end) is syntactic sugar for

    :wed-1 rdfx:occurrenceOf << :liz :spouse :richard >> .

but I’m still not sure :-/ 

>>>  PREFIX : <https://schema.org/>
>>>  << <#bp23> | <book> :datePublished "2023" >> .
>>>  << <#bp23> | <book> :publisher <X> >> .
>>>  <#bp23> a :PublicationEvent ;
>>>      :location <London> .
>> 
>> You do introduce a graph here. Just a slight change of the syntax and it would be "official":
>> 
>> PREFIX : <https://schema.org/>
>>    << 
>>       <#bp23> | 
>>       <book> :datePublished "2023". 
>>       <book> :publisher <X>.
>>    <#bp23> a :PublicationEvent ;
>>                    :location <London> .
>> 
>> It's also obvious how this would syntactically map to RDF named graphs. Of course a merely syntactic mapping would lose the semantics. But that again could be added via a description (ideally using a standard vocabulary, as the RDF 1.1 WG envisioned in the Note on dataset semantics).
> 
> In this specific example it could work, since the triples in the graph {<book> :datePublished “2023”. ,  <book> :publisher <X>.} seen as “definite descriptions” {“the publication of <book> in 2023”, “the publication of <book> by <X>”} can actually refer to the same event. In general this wouldn’t be true for an arbitrary graph.
> This emphasises that generalising to graphs requires to look extensively at their use cases first in order to better understand the appropriate semantics.

You would have to provide a crystal clear definition of "definite descriptions" to make sure that people will not misuse this mechanism. That seems like a futile endeavour to me. Generalising the mechanism right away and providing appropriate caveats is probably much safer, and good enough for most use cases. 

The usage of graphs is just too manifold to hope to arrive at a complete picture of use cases in any useful detail. The most superficial characterization will have to do: they group statements into entities that refer to something distinct from (although of course related to) the statements they group. 

BTW IMHO this is a better description than yours. You claim a common "definite description" but still a difference between those two statements - and that just doesn’t make much sense to me. Those two statements describe a book publishing and that may seem like a pretty strong commonality. But any other two statements can have *some* commonality and who are we to decide (and how are we to unambiguously explain that decision) what constitutes a strong enough commonality? That’s impossible, and futile, and any attempt to it would probably even feel presumptuous to users.

Since grouping of statements is such a basic KR activity there can’t be any reasonable doubt that whatever we provide will be used to represent graphs. So we *really* better account for it right away before real world usage establishes uncompatible practices (like eg interpreting two occurrences with the same name as merge or sameAs). 

>>> To better see that, you can verbalise the triple term (the truth-bearer) as a definite description:
>>> you are saying that “the publication of <book> in 2023” and “the publication of <book> by <X>” in that graph snippet both refer indeed to a single publication event located in London; and, clearly, those triple terms (as definite descriptions) could refer to something different in other parts of your graph.
>> 
>> Just to clarify that I understand correctly: by "triple terms (as definite descriptions)" you mean the triple terms themselfes, e. g. used "stand alone" or in other combinations with other triple terms in other (in this syntax somehow "virtual", only denoted by the occurrence identifier) graph occurrences?
> 
> Yes.
> 
>> All this confirms me in my intuition that an extension of the current proposal towards graph terms/occurrences is both unavoidable and straightforward.
> 
> Mmmhhh :-)

;-)
T.

> cheers
> —e.
> 


>

Received on Friday, 5 January 2024 14:24:47 UTC