Re: Presentation about reifiers and graphs

Le 17/09/2024 à 06:21, Franconi Enrico a écrit :
> On 17 Sep 2024, at 05:20, Thomas Lörtsch <tl@rat.io> wrote:
>>> Slides 19 and 20 ==> Uh? It is not clear at all what you want to say. 
>>> It i probably enough to say: a set of triples associated to the same 
>>> identifier can not be seen as a named graph, since the scope of the 
>>> bnodes in that graph is not limited within that graph but it spans 
>>> the whole main graph.
>>
>> Not necessarily, as "Blank nodes *_can_* be shared between graphs in 
>> an RDF dataset." [0]
>> [0] https://www.w3.org/TR/rdf11-concepts/#section-dataset
> 
>  1. In the above “encoding” of a named graph, blank nodes *_MUST_* be
>     shared between graphs in an RDF dataset.

I would say "can" is the right word here. There are RDF datasets in 
which the named graphs do not have any blank nodes in common, therefore 
they do not share bnodes. You probably think of "sharing" in a different 
sense.

>  2. Datasets have two formalisations of semantics [1]: default graph as
>     union or as merge, characterising exactly the sharing or not sharing
>     of bnodes.

Please don't use [1] as if it had any authority. It's just a note about 
the many ways of formalising semantics of RDF datasets. Among possible 
formalisations, there are those where the default graph is meant to be 
interpreted as the union of the graphs in the named graphs, those where 
it is the merge instead, and those where it is neither.

Also, sharing or not sharing bnodes is a question of syntax. Sec.3.2 of 
[1] considers that bnodes can be shared among named graphs. Yet, with 
the same bnodes in multiple named graphs, one can still interpret a 
dataset as the merge.

In case it is not entirely clear, take b a bnode, and u1, u2 two 
distinct IRIs. G1 = {(b,u1,u1)} and G2 = {(b,u2,u2)} are different RDF 
graphs that share the same bnode. If G1 is true, then there is a 
resource related to what u1 denotes. If G2 is true, then there is a 
resource related to what u2 denotes. Both G1 and G2 could be true with 
two distinct resources related to what u1 and u2 denote respectively. 
Said differently, {G1,G2} being true is equivalent to 
{(b1,u1,u1),(b2,u2,u2)} being true, with b1 and b2 different bnodes. 
This is merge.

But one could also think of G1 and G2 as pieces of a larger graph G = 
{(b,u1,u1),(b,u2,u2)}, thus interpreting them together as their union.

Then, when you put the graphs G1 and G2 inside named graphs, there is no 
standard to how the collection of named graphs must be interpreted, 
neither as union, nor as merge, and there are many more options that do 
not require any of the two choices, as exemplified in [1].

--AZ

>  3. In Trig, a blank node label represents the same blank node
>     throughout the TriG unique document, i.e., blank nodes sharing the
>     same label in differently labeled graph statements are considered to
>     be the same blank node. >  4. In general, when named graphs in a dataset have different origin,
>     you cannot assume that the same blank node label in different named
>     graphs represents the same blank node.
> 
> 
> *My conclusion*: these things need to be discussed, and we haven’t done 
> it yet, and therefore I’d restrain to say that we can represent named 
> graphs.
> 
> —e.
> 
> [1] 
> https://www.w3.org/TR/2014/NOTE-rdf11-datasets-20140225/#default-graph-as-union-or-as-merge <https://www.w3.org/TR/2014/NOTE-rdf11-datasets-20140225/#default-graph-as-union-or-as-merge>
> 
> 

-- 
Antoine Zimmermann
École des Mines de Saint-Étienne
158 cours Fauriel
CS 62362
42023 Saint-Étienne Cedex 2
France
Tél:+33(0)4 77 49 97 02
https://www.emse.fr/~zimmermann/

Received on Tuesday, 17 September 2024 06:20:12 UTC