On Tue, 29 May 2001 22:50:48 Horst Herb wrote:
>> Ontologies are a very modern topic and a very important one, since it is
>> probably the key to medical decision support.
>
>Why probably? I would say definitely.
I agree.
>> Should the ontology be an initial component for a project (efficient but
>> very managed option - Odyssee choice), or should it be built by the
>> community as an initial stage of the project (freedom and entropy - OIO
>> option).
A little clarification here. My proposal is that the ontology be built _continuously_
and with minimum overhead by community members (through software tools).
It starts from day one of the project and continues to evolve. It always remains a
work-in-progress, is used while "incomplete", and always supports growth and
modification. It does not transition from "initial" to "non-initial" stages.
>Although I can see where Andrew comes from, I don't believe that it will
>work with too much ambiguity (entropy or whatever you want to call it).
The problem is that ambiguity cannot be "mandated" away. Human languages evolve (with
human understanding?). Any ontology that assumes full compliance and proper use by
"humans" is just 'sweeping the dirt under the rug'.
What I am trying to say is that "ambiguity" does not come from ontology or the tools
that facilitate the "description" and automated processing of ontology. "Ambiguity"
comes from scientific advances, human creativity/ignorance, etc.
>If I want to check for things like drug-disease interactions in decision
>support, I would need true AI if I would allow any ambiguities.
Exactly my point. One way to look at this problem is to _assume_ that we have an
option to "not allow" ambiguities. I disagree.
For drug-drug, drug-disease, and other uses of the ontology, my proposed approach does
not necessarity lead to more or less ambiguity than Philippe's approach. In fact, my
approach does not preclude the use/incorporation of existing ontologies into the
proposed "flexible" framework.
I would further propose that Philippe's "pre-built" approach is potentially
synergistic with my "evolving" approach. Starting with a "pre-build" framework may
make it easier to stitch together pieces of the "evolving" ontology.
>Thus, I believe that we depend in an absolute way on clearly defined
>ontologies if we want to achieve *safe* computer asisted decision support in
>medicine, and we depend on it to in order to allow meaningful globalized
>data mining.
"Clearly defined" is a goal and a moving target rather than a *feature* of any
ontology. Similarly, *safe* decision support will involve incremental development.
Before we hand-over "decision making" to computers, we will have to "teach" them to
make decisions. This implies that the computers must have the capability to "learn".
That, is what I am proposing by my "flexible" ontology.
>I further believe that commercial ontologies (royalty ontologies) will never
>be accepted globally - thus being worthless for the "greater picture". I
>have great respect for OpenGalen, and I realize that I have to have a closer
>look at Odysee for that matter.
In my prior message (tagged with [ontologies]), I asked about the ability to
incorporate new terms (I used the example of a newly discovered disease). IMHO, that
is the most challenging and crucial feature of any tool that attempts to
describe/construct ontologies.
Best regards,
Andrew
---
Andrew P. Ho, M.D.
OIO: Open Infrastructure for Outcomes
TxOutcome.Org (hosting OIO Library #1)
Assistant Clinical Professor
Department of Psychiatry, Harbor-UCLA Medical Center
University of California, Los Angeles
Join 18 million Eudora users by signing up for a free Eudora Web-Mail account at
http://www.eudoramail.com
