Re: use/mention and reification

I think I'm following this discussion, but on the other hand maybe I'm
not, so I've inserted some comments/questions below.  Part of my problem
is that the examples seem to be getting further and further away from
literal (if you'll excuse the term) RDF reification (using triples or
graphs instead of English might help), so it may not be as easy as it
might be to apply the examples to the specific problems we have to
solve.    

Jan Grant wrote:
> 
> [snipped for brevity to get to the crux of the issue]
> 
> On Tue, 22 Jan 2002, Dan Brickley wrote:
> 
> > Can we try a variation on this example instead, please? For those who get
> > the cultural reference ('clark kent' and 'superman' being two names for
> > the same thing, although Lois doesn't know this) it seems to draw out the
> > issues nicely. Borrowing from
> >  http://plato.stanford.edu/entries/prop-attitude-reports/index.html#amb :
> >
> >       lois accepts "Superman is Strong"
> >       (lois does not accept "Clark Kent is Strong")
> >
> > > where "said" means, "emitted the following symbols"*, where I prefer
> > (we could fuss over quite what 'emitted the following symbols' amounts to,
> > but its good enough for me. i'm using 'accepts' instead of 'said' but the
> > example still works, I think)
> >
> > >     jan said that mary had a little lamb

One problem I'm having is that a lot of the meaning in examples like
this one comes from the meaning you read into expressions like "said
that" (I found "emitted the following symbols" to be much more precise,
in this connection).  For example, one way to read the example 

      jan said that mary had a little lamb

is that this is what jan *meant*, even though he actually may have said
(the words he emitted may have been) something else.  If that's so, I've
then got to worry about whether whoever is telling me what jan meant not
only heard what he actually said properly, but also shares jan's
interpretation of what that was intended to mean. With that (possibly
wrong) understanding, I (think I) share Dan's concern about "information
loss".  At least if I knew what jan literally said (the words he
emitted), I'd have a concrete fact, and I could separate the question of
whether I agree with jan about what those words refer to.  

I also wonder whether using different properties might clarify things
(or at least cast a different light on them).  For example, suppose I
use "has date of", as the property, 01/22/2002 as the object, and for
subjects I want to choose things so I can distinguish 

a.  the date the statement "mary had a little lamb" was made (forget who
said it, for this example)
b.  the date mary had a little lamb (assuming "mary" doesn't refer to a
sheep, she was participating in a bizarre genetic experiment)

What should the subjects be?

In dealing with RDF reification, we have to assume that people will use
all sorts of properties (like "said" in the original example) to express
relationships between something (like jan) and a reified statement.  We
(RDF Core) can't control what people intend by the various properties
they define;  what we do need to be clear about, though, is what the
reified statement means.

> >
> >       lois accepts Superman is Strong
> >       (lois does not accept Clark Kent is Strong)
> >
> > [substituting co-referring terms...]
> >
> >       lois accepts Clark Kent is Strong
> >       (lois does not accept Superman is Strong)

The problem I have with *this* example is that these aren't co-referring
terms *to lois* (a point Dan made in his original presentation of the
example, but then got away from);  in lois's "context" (or assignment of
meanings to names) Superman and Clark Kent are different people.  I
don't think the problem here is some kind of use/mention confusion, it's
that lois and we readers of DC Comics <http://www.dccomics.com/>
associate different semantics with Clark Kent and Superman (mostly with
Clark Kent).  Morever, we readers of DC Comics have to be aware of this
contextual difference in order to understand what's going on in the
stories.  I don't think we'd better expect any specific version of
reification to eliminate these situations.  

> You _do not_ need to quote URI labels on resources to make this
> distinction. <foo:superman> and <foo:clarkKent> are labels on resources;
> these labels are preserved using my version of reification; and the fact
> that an interpretation means they denote "the same thing" isn't a
> problem at all.
> 
> > So I'm claiming that more often the 'meaning not symbols' (de re) approach
> > to rdf reification results in unacceptable information loss
> 
> I think it's clear now that it doesn't. Meaning comes from the
> application of an interpretation function, which is where "clark kent"
> and "superman" collide. Quoting URIs makes no difference here.
> 
> > Both forms have their uses. But we can go from a de dicto ascription to de
> > re safely (I think...[1]), but not back the other way. Given that
> > asymetry, chosing the preserve-the-symbols approach seems pretty
> > attractive.
> 
> Preserve the resources, maybe, but turning URI labels on them into
> strings isn't necessary here.

Is the question really whether URIs should be quoted in reification (I
assumes this means in reification syntax), or what the interpretation of
a reified statement (that is, the actual syntax or graph of the reified
statement) is supposed to mean?

--Frank


-- 
Frank Manola                   The MITRE Corporation
202 Burlington Road, MS A345   Bedford, MA 01730-1420
mailto:fmanola@mitre.org       voice: 781-271-8147   FAX: 781-271-8752

Received on Tuesday, 22 January 2002 16:14:10 UTC