Re: Possibility of an LPG-profile

On 21/08/2024 17:02, Lassila, Ora wrote:
>
> I have been thinking a lot what I would like to see in an 
> “LPG-profile”. Here are my current thoughts:
>
> The general question, of course, is this: “Are profiles harmful for 
> RDF-star, and if they are, why?” In other words, if we define 
> profiles, what will blow up downstream?
>
> A minimal “LPG-profile” (for the lack of better name) would be a 
> restriction on the current baseline model of RDF-star that would make 
> the rdf:reifies -property (or whatever name we ultimately choose for 
> it) functional. In other words, it would ensure that a reifier 
> (identifier) only reifies a single triple. This allows us to treat 
> reifiers as LPG edges, and statements about reifiers (i.e., statements 
> about “triple occurrences”) would then correspond to LPG edge 
> properties. RDF is more expressive, though, allowing the objects of 
> these statements to be nodes in the RDF graph, not just literals like 
> in LPGs. This would be fine.
>
> My questions about any of this are at least the following:
>
> 1 – Is this a syntactic or a semantic restriction?
>

Simple entailment RDF doesn't express e.g. cardinalities,  so I was 
imaging a profile to have syntactic restrictions c.f. SHACL.

The target domain does not have semantics. We have to consider whether 
higher-level semantics imply anything about the target domain.

> 2 – If we do this, or some other profile-based restriction, what are 
> the consequences for semantics (as defined in the baseline)?
>
> 3 – Is it still possible to define entailment regimes (beyond simple 
> entailment) on top of this (RDFS & OWL, namely)? That is, can the 
> restriction be violated by way of entailments?
>
> 4 – From the implementation standpoint, are there aspects of RDF-star 
> we have not yet considered? For example, does allowing IRIs as 
> reifiers present a burden for implementers (say, considering scenarios 
> where a triple store learns after the fact that the subject of some 
> statements is indeed a reifier)? Discussing implementation details of 
> upcoming commercial products is not easy, of course. Since an 
> LPG-profile would be (potentially) interesting to people also 
> implementing LPG support, their implementation choices may not be the 
> same as those just implementing RDF. Your mileage may vary.
>
> Sorry I did not get this email out earlier. I am hoping we can discuss 
> the general idea and possibility of profiles tomorrow.
>
> Ora
>
> -- 
>
> Dr. Ora Lassila
>
> Principal Technologist, Amazon Neptune
>