ontolog-forum
[Top] [All Lists]

[health-ont] Re: [ontolog-forum] proceedings from the Thu 2005-01-06 Co

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
Cc: "[health-ont]" <health-ont@xxxxxxxxxxxxxxxx>
From: "Peter P. Yim" <peter.yim@xxxxxxxx>
Date: Thu, 06 Jan 2005 16:40:57 -0800
Message-id: <41DDDA99.9070709@xxxxxxxx>
 > In addition to the above, Nicolas Rouquette took some notes of the 
call proceedings that he will be
 > sending in, along with other relevant artifacts which he will be 
sharing with us too. -Thanks, Nicolas    (01)

Here're the notes from Nicolas: 
(I actually didn't realize Nicolas got them off to my inbox even before 
I had sent out my earlier report ... Sorry, Nicolas! -ppy)    (02)

About Conrad Bock's upcoming talk on January 27th:    (03)

Peter mentioned that Conrad worked on PSL at NIST.    (04)

PSL is a really nice ontology because it's design is very modular,
meaning that one can use only the relevant pieces of PSL that are
relevant. This is especially important to avoid ontology "bloat"
due to upper ontologies where a small fraction of their content
is actually relevant to a specific extension or application.    (05)

Modularity in ontology is often cited as a desirable attribute.
Recently, I found further evidence to support a stronger claim that
modularity can be seen as an essential property towards 
the cost-effective ontological development. Only a fraction of
a large upper ontology can be relevant to a specific extension (e.g., 
middle/lower ontology)
or to a specific application. The irrelevant fraction of the upper ontology
adds unecessary burden, complexity that amounts to "dead" ontology weight.    (06)


Nataly's article [1] mentions an experiment where OWL-S and OWL-WS
were mapped and that mapping allowed Protege and its plugins to
construct from an extension originally developed for OWL-S into a new
extension for OWL-WS. Due to high development costs involved in
mapping and extending ontologies, the above experiment aludes to the
prospect of "ontological extensions for free" made possible by the
formal ontology development. (There is in fact a notion of "theorems for free"
in category theory that, coupled with the notion of Galois connections leads
to some very powerful deductive mathematics -- see Backhouse & Backhouse)    (07)

[1] Semantic integration: a survey of ontology-based approaches. Natalya Noy, 
ACM SIGMOD Record, vol 33, no. 4, December 2004    (08)

================================================================================    (09)

a) ontology candidates for an "upper" ontology    (010)

- SUMO
- HL7
- OpenCYC
- ...    (011)

b) requirements for choosing an upper ontology    (012)

 b1) what is the content of this ontology?    (013)

- ontology as a normative definition of concepts or as a communication 
interchange "format"?
- static content [2]
- dynamic content [2]
- intention content [2]
- social content [2]    (014)

[2] Ontologies for Knowledge Management: An Information Systems Perspective
Igor Jurisica, John Mylopoulos, Eric Yu
Knowledge and Information Systems, vol 6, p. 380-401, 2004.    (015)


 b2) how can we assess this ontology?    (016)

- precision
- internal consistency
- axiomatic definition
- open nature of the formal language used to define the semantics of the 
ontology
- extent of the formal definition of the ontology concepts    (017)

 b3) extensibility    (018)

- formal rigor in which one can define an extension of an upper ontology for 
practical purposes
- dependence on proprietary languages/semantics/tools hinders open-source    (019)

  extensibility, distribution, accessibility and use    (020)

--
Notes from the 2005.01.06 Ontolog conference call 
by Nicolas Rouquette <nicolas.rouquette@xxxxxxxxxxxx>
2005.01.06_12:29 PST    (021)



 _________________________________________________________________
Msg Archives: http://ontolog.cim3.net/forum/health-ont/
Community Files: http://ontolog.cim3.net/file/work/health-ont/NHIN-RFI/
To  Post: mailto:health-ont@xxxxxxxxxxxxxxxx
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?NhinRfi    (022)
<Prev in Thread] Current Thread [Next in Thread>