RE: SUO: RE: Topic Maps
John
Thanks for your reply. What I would like to do next is to provide a more
explicit statement of the ways in which I believe that XTM might of utility
in the endeavours of the group and in some way validate it's ability to
support those possible roles.
Chris Angus
-----Original Message-----
From: owner-standard-upper-ontology@ieee.org
[mailto:owner-standard-upper-ontology@ieee.org]On Behalf Of John F. Sowa
Sent: 17 February 2001 17:45
To: standard-upper-ontology@ieee.org
Subject: RE: SUO: RE: Topic Maps
Chris,
I believe that we more or less agree on the basic principles.
>I would question whether every notation has to be explicitly mapped to FOL,
>it might be sufficient to map to one or more other notations whose
behaviour
>with respect to FOL are well known and understood.
Yes. The main point is to identify how any proposed notation
relates to the ones that already exist. For programming
languages, the ability to simulate a Turing machine is a key
test that separates the general-purpose languages from the
very specialized ones. And once you prove that langauge A
can simulate a Turing machine, it is only necessary to show
that another language B can simulate A.
>I would also argue that describing a modelling notation in terms of FOL is
>only a part of the picture - the weakness of many notations, it seems to
me,
>is that the mapping between the model and the world is not sufficiently
>defined (which I guess is often but not always an ontological problem).
Yes, the ontology is often the most important part. KIF and
CGs have no built-in ontology, but most common CG applications
have a distinct bias toward the ontology that I presented in
Appendix B of my 1984 book. The same is true of most other
languages, and many of them fail to identity what is part of
the notation, what is part of the common applications, and
what is optional.
>It is certainly true that the XTM (the XML topic map notation) 1.0
>specification does not contain a formal mapping to FOL. There are moves
>afoot however to improve the rigor of the processing model and to encode it
>in a number of representations. That should make the mapping to FOL pretty
>explicit. The original topic map bias may well owe much to the world of
>publishing but there is now a greater representation from people who would
>like to see more rigor introduced - hence a greater interest in the
>underlying conceptual model and the role of ontology/ontologies.
>That said, my first question, in terms of the use of topic maps, is "would
>they bring utility?". If the answer to that question is "no" then their
>relationship to FOL is irrelevant. If the answer is "yes" or "maybe" then
>the relationship to FOL becomes a substantive issue.
I would say "maybe", and a clearer understanding of how they
relate to FOL would help us decide how, where, and when to
use them in addition to or instead of some other notations.
John Sowa