Re: blank nodes (once again)

On Mon, Mar 14, 2011 at 5:56 PM, Ivan Shmakov <ivan@main.uusia.org> wrote:

>        I guess I'm close to become an annoyance, but I'm going to ask
>        yet another blank node question anyway.
>
>        Suppose that I have the following graph:
>
> :joe :has [ a :dog ]
>
>        And suppose that I also have the following one as well:
>
> :joe :has [ a :dog ]
>
>        From the previous discussion I've learned that there're
>        different opinions on whether to consider these graphs, which
>        share exactly the same representation, same or different.

This looks close to my addition to Michael Schneider's addition the
bnode-questionare[1]. The way the question was phrased was certainly
influences by my opinion that current semantics for bnodes is sound, that it
is pointless to redefine bnodes to work the same as URIs and that
differentiating between named and anonymous graphs is a rather weird rise of
complexity by mixing in another buzz.

So according to current semantics the two isomorphic graphs above express
the same content and mutually entail each other so the are semantically
equivalent. According to RDF semantics blank nodes have no intrinsic names,
yet it mentions the possibility of skolemization where blank node are seen
as purely syntactical names. So one could argue that that the above could be
two serializtaions of syntactically different graphs. Personally I don't
think that this makes much sense, I think two graphs should be considered
equals if they are isomorphism and equivalent (i.e. express the same
content) if they lean versions are isomorphic.


> In
>        particular, there's a position that these graphs are different,
>        unless they're named the same.  (Somehow, I feel that graph
>        naming should be considered tangential to the knowledge it
>        represents, but I've noted to myself that there's a different
>        opinion.)
>
I completely agree with you.

If you assign a name two each of them, they both have two names (as they are
the same graph) [2].


>
>        But my question is itself tangential to the equivalence of these
>        graphs.  Instead, I wonder, if I've assimilated this above
>        representation into an RDF store, and going to assimilate its
>        exact twin again, does this later assimilation change the
>        /knowledge/ contained within such a store, or not?
>
No. The /knowledge/ will remain the same, even though the store might no
longer by lean after adding the two identical graphs.


>
>        To speak it differently, I've never heard of Joe, and
>        (unexpectedly) received a bit of information that speaks: Joe
>        has a dog.  Now, I receive another bit, that says exactly the
>        same.
>
>        I'm quite certain that after I've received the first bit I now
>        have a bit more knowledge about the World.  However, I'm not so
>        sure that the second bit gives me any more knowledge, since I
>        still have no rational means to tell, whether the dog I'm told
>        of this time is the same or different to the one about which
>        I've already known.
>
You don't know anything more, except maybe that two sources assert that Joe
has a dog.

I think bnodes are a powerful bit of RDF and I hope the discussion about
them will yield to better documentation on when and how to best use them
rather than a change of semantics. In the long term I think the cost of
having too many names for the same entities[3] are much higher than those of
handling the complexity of bnodes and leanification.


Cheers,
Reto

1. http://lists.w3.org/Archives/Public/semantic-web/2011Mar/0108.html
2. In Apache Clerezza immutable graph objects are equals if the are
isomorphic, thus TcManager returns a set of names for a Graph (
http://incubator.apache.org/clerezza/mvn-site/org.apache.clerezza.rdf.core/apidocs/org/apache/clerezza/rdf/core/access/TcManager.html
)
3. See last paragraph of:
http://lists.w3.org/Archives/Public/semantic-web/2008Jan/0118.html