Re: silly question about rdf:about

From: "Dave Beckett" <dave.beckett@bristol.ac.uk>

> >>>Seth Russell said:
> > From: "Dave Beckett" <dave.beckett@bristol.ac.uk>
> >
> > > >>>Seth Russell said:
> > > And while were at that, why not invent another useful property ...
> > > something to mean 'preferred human friendly name' ... this would be
> > > like a cyc constant, a kif term, or a rdf:label.  The thing that is
> > > different between it and rdf:about (aside from the fact that it would
> > > be human user friendly) is that it can change (be renamed) from time
> > > to time and from system to system.  But in any given system at any
> > > given time  it would be unique.  For a moment let's just call this new
> > > term ':named'.   Nodes so named internally would be tied to
> > > URI like this:
> > >     [:uri <http://foo/#Dog>;
> > >       :named  "Doggie"].

> > {Seth likes Chocolate}.
>
> {} brackets now, was [] above.  Is this significant?

Maybe, maybe not ... depends on the application implementation.  If we know
that 'named' is a unambiguous property, which will always be asserted of a
token where nothing is yet known about it, then we could have said {Doggie
uri "http://foo/#Dog" }  just as well.  I didnt want to assume that you
shared all of those assumptions when I originally wrote the above []
statement.  In other words, there is an application, in which the []
sentence is entailed by the {} sentence.

> > > However, rdfs:label already exists, so why not just use that?
>
> Thanks for snipping all my reply and leaving just one line.

You're welcome .. I wanted to focus on what i felt was important in your
message and on that to which what I had legitimate responses.   What I
snipped I either agreed with (in that your corrections were legitimate) or
that I felt were totally irrelivant to my train of thought. Sorry you felt
slighted.

> > Because rdfs:label is not guaranteed to be a unique name for a node
within
> > one system (or application memory storage) .. as specified above [1] ...
yet
> > it can come from the outside world and cannot be ignored.  This is not
too
> > unusual of a concept, CycL does it.  Such an internal name property is
> > useful, because then you can just scribble something to your system
without
> > haveing to remember namespaces, URI or even include all those annoying
> > colons.  For example:
>
> I see, so you want a unique name.  In that case, you need schema
> systems such as DAML that have such restrictions.

Yes, you need the ability to say that only one node can have that
property/object pair within some context ... daml: UnambiguousProperty does
that.   The minimum you need to do, is to implement the rule {if P is an
UnambiguousProperty, then if P(x, y) and P(z, y) then x=z.} for each such
property.

>This isn't going
> to get into the core RDF schema at this time.

Doesn't need to ... this is an internal application implementation thing.  I
dont see what is gained by the W3C putting anything about it in the
standards.   Do you?

> > {Seth uri "http://robustai.net/~seth#ThePerson".
> >   likes uri "http://robustai.net/verbs/#likes" .
> >   Chocolate uri "http://robustai.net/EnglishNouns#Chocolate" .}
>
> This looks like reification into N3 contexts/formulas.

It's just N3 with the redundant colons stripped and uses quotes to isolate
the uri strings ... I call it Semenglish (as you probably know) , see [1]
[2].  I don't like to type a lot of unnecessary punctuation.  I notice that
CycL and KIF uses roughly the same technique.  Since here I am talking about
an isolated application, the same word used in different namespaces can be
easily disambiguified within the system ... obviously only one such synomym
may be permitted to have the 'named' property.

[1] http://robustai.net/mentography/semenglish.html
[2] http://robustai.net/mentography/SyntaxSemenglish7.gif
... work in progress .. everything might not be consistent ... you know,
kind of like cwm.

>So this is
> again beyond-RDF, and interesting but not something happening in the
> core RDF work at this time.

As above, doesn't need to be in the standards at this time.

> Are 'Seth' 'uri' etc URIs, bnodes?

They are like CycLConstant (s) .. except that we don't need no '#$' ...
everything that is not punctuation, or enclosed in quotes is by default a
constant.  Kind of like KIF, huh?

http://opencyc250.homelinux.org:3603/cg?cb-cf&5266

>And now you have '.' inside the {}s

Doesn't matter ...  there exists a sufficiently fast parser such that the
syntax would be unambiguous should the trailing period be inside or outside
of the {}.

> Please use a valid N3 or RDF/XML syntax

Request denied.

> >  ...  just translate the ':named' to the URI  and output it with a
> > 'rdf:about' .. not exactly rocket science. The obvious advantage is that
we
> > have eliminated the kind of pedantic problems which you have warned me
about
> > above ... disadvantages ... well I can't think of any ... can you?
>
> There is no :named in the second examples.

I assume, by the "second examples", you mean: {Seth likes Chocolate}.   Well
that statement entails: [ named "Seth"] , [named Chocolate] and [named
likes] within the application of which I speak.

> And if you want to talk
> about atranslations please show the translation in legal RDF/XML such
> as some that validates with http://www.w3.org/RDF/Validator/
> or N3 that works with Cwm.

Why, If I do, what will you or whatever do for me?  These are examples
scribbeled in emails, intelligent humans are the intended audience, not Cwm
or Jena.  Apparently you understood my  syntax enough to make insightful
remarks.

> The syntaxes you are using or inventing here aren't helping.

Says you.

> > Now obviously this is a API or application choice, it doesn't even need
to
> > be a standard in RDF.
>
> Not likely at this stage; but maybe at the WebONT language level
> or later RDF work.

I cannot compute the dialogue meaning of your response to my statement.

Seth Russell

  "If you don't want unicorns to exist,
    then don't talk about  them."
             Pat Hayes ~ 2002

Received on Monday, 8 April 2002 19:55:34 UTC