Re: class or instance?

From: "Steven Gollery" <sgollery@cadrc.calpoly.edu>

> Due to my lack of background in this area, I've always had a hard time
> understanding your mentographs. Can you point me to some sort of
introductory
> material?

There is a brief introduction at [1].  But really these are just labeled
directed graphs.  Every rectangle is a node.  What is put inside the node is
just to label the nodes for human understanding.  The only things that
exists in the database (or the model) are the arrows. You can always read
the database records directly off the graph, with one record being one
arrow.  The arrows are binary relations holding between the nodes and are
always labeled with another node.

[1] http://robustai.net/mentography/Mentography.html

Semantically you could consider that nodes represent something in the domain
of discourse.  The graphs themselves are just data structures (pure syntax)
and as such are language independent just like lisp (car, cdr) pairs.
Wherever I use a particular word from a known language, I have tried to
label it with the namespace to which it belongs (viz rdf:type) and then it
is assumed that the semantics of that arrow are exactly the semantics of the
word which labels it.   Sometimes I use shortcuts, which are mostly
described in [2].

[2]  http://robustai.net/mentography/MentographySemenglish.gif

> Also, I wonder if it is possible to express your mentograph in DAML.

Does not DAML have a fixed vocabulary with each build in term being a
property label of an RDF graph?  So anything that you can say in DAML, I
could put exactly in a mentograph using only arrows labeled with daml:* ;
but not necessarily the other way around.  As far as I know DAML does not
have any way to express context, and many of the things I talk about in a
mentograph have to do with the context of statings .. so I would find DAML
too restrictive and actually many of the entailments that I want to express
are better expressed in KIF which also can be translated into the arrows of
a mentograph.   In short when one writes in a mentograph (a labeled directed
graph), one has not made any commitment to a particular language.

> Not only
> would it help me to understand the graph a little better, but it would go
a
> long way to answering Huajun's original question.

Well I took Huanjun's original question to just need quantification to
clarify the issue.  In other words you could say whatever you wanted to say
about fish and instances of fish in FOPL with quantification and not
introduce any of the ambiguities that were his concern.  Could you not?

Seth Russell

---- wrt to the original question ----
 Seth Russell wrote:

 > From: "Huajun Zeng (Intl Staffing)" <i-hjzeng@microsoft.com>
>
> >   I am entangled when I try to use DAML to describe such a sentence:
> > "Tom likes fish". The following statement seems to work,
> >
> >    <p:person rdf:ID="Tom">
> >      <p:like rdf:resource="f:fish" />
> >    </p:person>
> >
> > but I found it only represents "Tom likes an intance of fish", not the
> > class of fishes. Futhermore, if another people, Jack, also likes fish
> > and is described as above, we cannot say Tom and Jack are similar.
> > Because they like their own fishes individually!
> >
> > It seems that if the class "fish" is also an instance, the problem is
> > resolved. But as I know, the occuring of type and subClassOf in
> > a same statement has no semantics.
>
> Your term 'like' has no semantics either.  I think you need to use a
schema
> where you can do quantification.  See for example:
> http://robustai.net/mentography/quantificationFish.gif
>
> Seth Russell

Received on Wednesday, 27 February 2002 14:36:41 UTC