On Mon, 21 Aug 2006, Donald Doherty wrote:
Creating explicit connections between all similar and/or identical entries
in two schemas is an arduous task that is impractical to do manually.
Actually, I recently had quite a good experience doing this same thing
with trying to align top-level concepts of a Patient Record ontology with
a 'foundational ontology'. There was a recent email [1] on this mailing
list about bio-zen that [2] caught my attention because it specifically
uses DOLCE [3] as a foundational ontology. I like the approach because it
causes you to 'ground' your semantics by thinking about them in the most
fundamental ways..
In my case, I was connecting my ontology with DOLCE
in order to formalize the domain independent terms, but in doing so any other ontology grounded
in this way also would be largely interoperable.
For example, for the concept of a diagnosis, I found DOLCE's
dol:information-realization class appropriate.
" Any physical entity that realizes an information object. "
So, the act that records the diagnosis is an dol:information-realization
which 'realizes' (dol:realize) a specific kind of information-object. The
information-object class tree is:
* information-object
* diagrammatic-object
* iconic-object
* linguistic-object
* word
* phoneme
dol:linguistic-object seemed most appropriate given it's definition:
" An information object ordered by (encoded according to) a language. "
Where the ordering is with respect to Pathology. So the act that
realizes a diagnosis (an angiogram, clinical exam, lab test, etc..) and
the diagnosis itself can be aligned with a foundational ontology in this
way.
Ofcourse, a foundational ontology is only as useful as it is well
organized and I've found DOLCE to be pretty methodical in it's
arrangement (but I found myself refering to CyC and SUMO occasionally
where DOLCE didn't have any decent definition for a concept)
Since it models at a very linguistic and congitive level, the nomenclature
might seem a bit odd at first (perdants, endurants, spacio-temporal
things, and what not) but I was able to find an appropriate place for most
of the top-level concepts I attempted to align.
Unfortunately, bio-zen was more oriented to concepts at the molecular
level than for my needs, but any connection made to DOLCE would
automatically orient themselves with bio-zen (or any other derivatives).
To contrast, compare, and reconcile at the data level I must make these
explicit connections.
When I do so mentally, I make a tentative connection, test it for validity,
and then maintain the connection or throw it out based on the result. (This
is also what is done physiologically through mechanisms like Long Term
Potentiation in the brain.)
Automate connections between synonyms, antonyms, etcetera based on local
analysis of elements.
As a matter of fact, DOLCE has a rather impressive mapping [4] to wordnet
1.6 that could (in theory - having only browsed this mapping) facilitates
some of this by virtue of wordnet's impressive lexical coverage.
[1] http://lists.w3.org/Archives/Public/public-semweb-lifesci/2006Jul/0088.html
[2] http://neuroscientific.net/index.php?id=download
[3] http://www.loa-cnr.it/DOLCE.html (DOLCE)
[4] http://www.loa-cnr.it/Files/OWN_description.txt (OntoWordNet)
Chimezie Ogbuji
Lead Systems Analyst
Thoracic and Cardiovascular Surgery
Cleveland Clinic Foundation
9500 Euclid Avenue/ W26
Cleveland, Ohio 44195
Office: (216)444-8593
[EMAIL PROTECTED]