Re: Use cases / examples needed for default behaviour with bnodes

Dear Enrico,

I disagree that the ":stated-by" or „:says" predicates are necessarily modal since we could understand „:says“ as some kind of literal quotation, it is not clear what the predicate is supposed to mean without further contexts, but that as a side-node.

In N3 the blank nodes in quoted triples are local and we have referential opacity and I can tell from many discussions I had that users are normally not very happy about that. So one argument would be that „it feels unnatural“, but that is highly subjective and I think I would even be happier with blank nodes in syntactical quoted triples being local. But that is just my opinion.

So, to make up some use case, let’s assume that we have some graph

:bob a :Client.
<<:bob :likes :sports>> :source <http://some-source>; :recorded "2021-07-07"^^xsd:date.

Now, let’s assume that this graph is some customer data which can only be shared anonymized, so it might become

_:x a :Client.
'<<_:x :likes :sports>>' :source <http://some-source>; :recorded "2021-07-07"^^xsd:date.
Now, we still would want that the data co-refers, i.e., that the two _:x refer to the same client „bob“ even though we do not want to show his exact URI (for example because that would map to his name).

You could now also question why we would want to have referential opacity in the first place for this specific example. To make up the whole example, I add that our source <http://some-source> does contain a lot of interesting information but the persons who created the data do not use all terms correctly and we have some extra knowledge that some terms are used correctly and others are not. We would apply external reasoning (for example by rules), to derive knowledge based on the concepts used.

So, that is a really hypothetical use case, but I hope that this will lead to an interesting discussion. I will try to defend transparent nodes for now, so please shoot ;)

Kind regards,
Dörthe



Am 09.04.2023 um 18:41 schrieb Franconi Enrico <franconi@inf.unibz.it<mailto:franconi@inf.unibz.it>>:

Hi,
I’d like to see some example / use case justifying that, without a TEP, a quoted triple has “transparent” bnodes.
All the examples I have seen are quoted triples in the context of predicates like :says or :stated-by, which are modal predicates and therefore fully transparent (i.e., TEP).
I believe that in a pure annotation case the bnodes should be non-transparent, a case which is not covered by the CG final report.
—e.