Re: Consolidating triple/edges -- named occurrence version

> On 5. Jan 2024, at 15:38, Olaf Hartig <olaf.hartig@liu.se> wrote:
> 
> On Fri, 2024-01-05 at 13:55 +0000, Franconi Enrico wrote:
>>> On 5 Jan 2024, at 14:46, Olaf Hartig <olaf.hartig@liu.se> wrote:
>>> 
>>> On Fri, 2024-01-05 at 10:58 +0000, Franconi Enrico wrote:
>>>> On 5 Jan 2024, at 11:42, Niklas Lindström <lindstream@gmail.com>
>>>> wrote:
>>>>> Of course, it has implications for how to define these
>>>>> occurrences
>>>>> (truth-makers, right [1]?), which we need to come to terms with
>>>>> together.
>>>>> For example, I think this makes sense:
>>>>> 
>>>>>    << :wed-1 | :liz :spouse :richard >> .
>>>>>    << :wed-1 | :richard :spouse :liz >> .
>>>>>    :wed-1 a :Marriage ;
>>>>>        :starts 1964 ;
>>>>>        :ends 1974 .
>>>>> 
>>>>> Would you agree?
>>>> 
>>>> 
>>>> Yes, it does.
>>> 
>>> Notice that this diverges quite a bit from the Property Graph
>>> perspective. PG folks would understand the IRI :wed-1 to be an
>>> identifier of an edge, and they would see two edges here (one from
>>> :liz to :richard and another one from :richard to :liz). Then, they
>>> would get confused because two different edges cannot have the same
>>> identifier.
>> 
>> I agree. However, as I observed yesterday, the identity of PG edges
>> should be encoded in RDF by the property names (emulating a singleton
>> property), and not by reification.
> 
> Mmh. From a PG perspective, the predicate of a triple---assuming
> triples with a non-literal object---is like the label of an edge. In
> this sense, what you are proposing would appear to PG folks as
> (mis)using the edge labels as means to identify/distinguish all the
> edges.

RDF has that inherent wobbliness that syntax refers to meaning and only meaning counts in the end. The benefit of all this wobbliness is entailments, and integration powers far beyond individual domains. IMO LPG people will have to live with that as otherwise it’s not RDF anymore (the reason why I always was against referential opacity as the default configuration).

>> Reification then helps to add properties to the uniquely identified
>> edges.
>> This is of course possible with the current semantics.
>> 
>>> Notice also that a semantics such as this would probably not be
>>> very
>>> useful for provenance use cases. I would assume that, in such use
>>> cases, it makes a difference whether the provenance annotation is
>>> about
>>> the triple (:liz, :source, :richard) or about the triple (:richard,
>>> :source, :liz).
>> 
>> 
>> I also agree here. Proper provenance use cases do need a more
>> “syntactical” approach and therefore some form of opacity.
>> However, the WG decided to go full steam to a fully transparent
>> (i.e., semantic) approach.
> 
> When did this happen? In the WG meeting before Christmas (which I
> didn't attend) or in any of the SemTF meetings?

In my understanding around that time in a semTF meeting (the last before Christmas or the one before that IIRC). It seemed like a general consensus to me, but not a really well discussed one  (and I expressed my reservations, but I’m not sure yet what I prefer)

Thomas

> Olaf
> 
> 
>> That’s why yesterday I proposed that we could study extensions of
>> this current approach to include optionally some form of opacity
>> (whatever this may mean…).
>> 
>> cheers
>> —e.
>> 
>

Received on Friday, 5 January 2024 15:01:05 UTC