Re: CG: Re: SUO: Negotiation Instead of Legislation
Chris, John,
> Even the upper level ontology published in your book, John (flawed as I may
> think it is), has proven incredibly useful as an analysis tool to people building
> OO, DB, KR, etc. ontologies. It has been better to have it around
> than to banish it because some people (like me) don't agree with it.
Are you are referring to the top-level ontology derived from the notions of firstness,
secondness and thirdness (http://www.bestweb.net/~sowa/ontology/toplevel.html)
or to his older one, based on the partitions {Situation, Entity}, {State, Process},
{Spatial_entity, Non_spatial_entity}, {Living_entity, Non_living_entity},
{Description_content, Description_instrument, Description_container}, ...?
I have found these last distinctions useful for
- signatures of basic relation types (spatial/temporal/thematic/... relation types);
- structuring the top-level ontology of WordNet and then (with additional
constraints) automatically discovered about 300 semantic errors in WordNet 1.7
(e.g. categories specializing exclusive categories, subsumption links used instead of
part-of links or member-of links, etc; I have corrected the problems manually);
- modelling various domains;
but I could not use Sowa's more recent ontology for any of these applications because
- anything can be viewed/classified as first, second or third, i.e. can be "considered
dependant" of 0, 1, 2 (or more) other objects, and hence
- there is no exclusion links between these top-level categories
Although any physical_entity may theoretically be seen as a process, some relations
seems to be only applicable to processes, others to physical_entities, others to
description content, others to description containers, ... and detecting the misuse
of these relations (or misuse of the connected objects' categories) is important.
The above cited exclusive categories permit to give signatures to relations and
to detect a lot of semantic errors in users' knowledge representations.
Although I am not against the firstness/secondness/thirdness based ontology
(I have merged it into WebKB-2's default KB, along with many others; it is not a
generalization nor a specialization of distinctions in other ontologies),
I still do not understand its use from a knowledge engineer perspective, nor from
a semantic error detection perspective.
John, Chris, I would be very interested to have details and examples on these two
aspects. (I have not found them in John's articles).
John, I also do not fully understand your recent emphasis against "legislated
ontologies" since the categories and definitions of various ontologies can be
inserted and interconnected into a same KB, without lexical nor semantic problems.
For example, in WebKB-2, a category identifier has for prefix the identifier of its
creator/source, and a graph is contextualized by its creator/source, thus permitting
various users/sources to have different opinions and re-use/relate/annotate other
users' categories and graphs. Thus, users may (but do not have to) re-use, specialize
or complete these "legislated ontologies". If they do, their knowledge is more
retrievable/re-usable, and "negotiations" might even be made automatic.
Philippe
P.S. Example of subsumption hierarchy of relation types coming from various sources:
http://www.webkb.org/bin/termSearch.cgi?term=pm%23relation&recursLink=%3E&hyperlinks
__________________________________________________________________________
Dr. Philippe Martin
Research Fellow at the CRC for Enterprise Distributed Systems Technology
(DSTC Pty Ltd; DSTC is W3C's Australian Office)
Address: Griffith Uni, School of I.T., PMB 50 GCMC, QLD 9726 Australia
Email: philippe.martin@gu.edu.au; Fax: +61 7 5594 8066
___________________________________________________________________________