Message 12608 of the SUO list

Subject: Re: SEMIS Bulletin
Date: Mon, 31 May 2004 17:17:04 +1000
From: Philippe Martin
In-reply-to: msg12252 by Frederick Kintanar



From Frederick Kintanar to John Sowa:
> You have in the past endorsed IFF, do you have specific ideas
> about how it can be used to relate your KR book upper ontology
> with OpenCyc or SUMO?  It wasn't used in MSO, or were
> there elements of PM's approach that already embodies some
> IFF elements?

The IFF is a "meta-ontological framework" providing techniques
for "manipulating ontologies as objects, relating ontologies
through morphisms, partitioning ontologies", ... thus extracting
common parts and generate a "lattice of theories" (with the
source ontologies at the bottom).

In WebKB-2, no such technique is implemented, and in the MSO
there is (so far) no relation between "ontologies as objects", only
relations between categories or statements belonging to different
ontologies (e.g. subtypeOf or exclusion relations between categories
and corrective_specialization relations between statements).
However, such relations are necessary for generating a correct
lattice of theories. Then, this lattice could be stored into the MSO
(with each ontology or theory represented by a category).
If such relations and category definitions do not exist, an
alternative for generating a lattice of theories is to use FCA-based
techniques on a collection of individuals indexed by the various
ontologies to be aligned/merged. However then, the result is only
true "for the given data". For example, in
http://suo.ieee.org/IFF/work-in-progress/IF/example.pdf
Figure 3 shows that "stream" specializes "riviere": this is not
true in general (as noted in the paper) but only for the few
considered instances of riviere and stream.
For this example, what would be stored in the MSO is:
water_flow > {river stream} {fleuve riviere};
  river = [water_flow *w, attribute: a large width],
        > {fleuve large_riviere},
        : Ohio_river;
    fleuve = [water_flow *w, object of:(a flowing,destination:a sea)],
           : Mississipi_river Rhone_river;
  stream = [water_flow *w, attribute: a small width],
         > small_riviere;
    small_riviere = [riviere *r, attribute: a small width],
                  : Captina_river Roubion_river;
  riviere = [water_flow *w, object of:(a flowing,destination:no sea)],
          > {small_riviere large_riviere};
    large_riviere = [riviere *r, attribute: a large width];


> It wasn't used in MSO, or were there elements of PM's approach
> that already embodies some IFF elements?

In case "elements" refer to basic categories such as "set",
the answer currently is: only if such categories are in the SUMO or
in DOLCE.


> relate your KR book upper ontology with OpenCyc or SUMO

John Sowa's top-level ontologies are in the MSO. There are some
connections with the SUMO (the SUMO re-uses some categories from
John Sowa). OpenCyc is not in the MSO.


Philippe