Re: Can we express propositonal attitudes twards reified statements in RDF ?

>Pat,
>
>The WG agreed:
>
>     "The group overwhelmingly, unanimously supports that
>      we should, in  principle, focus on addressing the
>      provenance use-case." [1]
>
>You said:
>
>        "Well, its (literally) impossible to give a coherent interpretation
>         of reification which satisfies everyone. We had to choose one,
>          and we chose the one that seemed to support the existing use
>          cases that people felt strongly about. "  [2]
>
>So can I assume that the subsequent choices of the WG did in fact 
>support the provenance use-case?

Yes.

>Then you say:
>
>      "In the present set-up, the reified triple is required to mean
>        what it would mean if you de-reified it. It refers to the proposition,
>        not to the surface syntax. "[3]
>
>In the light of the above wouldn't it be more accurate to say that 
>the reified node refers to the stating of the proposition, and not 
>the proposition itself ?

Yes, it would. There are two dimensions here, in fact, which are kind 
of orthogonal:

stating/statement  (what the subject of rdf:subject/object et. refer to)
de re/de dicto (what the object of the reification vocabulary refers to)

We chose the stating/de re combination. What this means, in brief, is 
this: if I write for example

aaa rdf:subject bbb .

then aaa refers to a stating (not a statement) and bbb does not refer 
to the syntactic subject of the triple in the stating, which might be 
a bnode or a uriref (de dicto)., but rather to whatever that subject 
refers to (de re).  And similarly for rdf:object, etc.

The stating choice was largely motivated by the provenance use case, 
and that is what your citation [1] refers to. The other choice was 
also based on consideration of use cases, but I do not recall clearly 
what they were.

What this combination allows one to do is to state a relationship 
between a particular document and some entity which the RDF triples 
in the document are talking about. This was what was needed for the 
provenance use case, as I recall.

>Now we all know that we cannot  substitute in a referentially opaque 
>context [4]. 
>I don't follow the reasoning that gets us from there to your statement:
>
>    "In a nutshell, :thinks isn't a relationship between
>      an agent and an RDF reification, so it can't be an RDF property. "[5]
>
>Could you elaborate that reasoning for me?

Well, the choice of the de re semantics means that it is possible, in 
effect, to substitute into a reification context. It isn't 
technically possible in RDF since there is no RDF equality, but if 
you put together the intended semantics for RDF reification and those 
of say OWL, then  the combination
...
aaa rdf:subject bbb .
bbb owl:sameIndividualAs ccc .

entails

aaa rdf:subject ccc .

So reifications are not opaque; so they are not strictly a suitable 
choice for representing an opaque context, such as the object of 
:thinks: or :believes, etc.

However, as I noted in an earlier reply to Tim, there are probably 
ways to use terms like :thinks in a referentially transparent 
framework and hack round the rare cases where this might give 
problems, like the Superman case. Whereas the choice of the de dicto 
semantics would have made the logic very much more complicated and 
much harder to use in the 'normal' case where all the agents involved 
share a common understanding of the names in use.

Pat

>
>[1] http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Feb/0263.html
>[2] http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0237.html
>[3] http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0237.html
>[4] http://gncurtis.home.texas.net/illisubs.html
>[5] http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0229.html
>
>
>Seth Russell
>http://robustai.net


-- 
---------------------------------------------------------------------
IHMC					(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola              			(850)202 4440   fax
FL 32501           				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam@ai.uwf.edu   for spam

Received on Wednesday, 12 February 2003 17:26:58 UTC