Hi,
I would tend to agree with Matt on this one. An RDF version of the
UMLS knowledge sources would be seem to be very useful - at least for
bioinformatics research purposes - without the benefits of the
"correct" OWL ontology with which to describe the relationships
included.
Though there is no doubt that some ontologies are better than others
and that there is clear value in investing in building good ones
(e.g. a good OWL representation of the UMLS S.N.), the idea that, as
a community, we will ever reach a consensus for the one right way to
interpret the relationships in a resource as broad as the UMLS seems
unlikely. It seems to me that if we really want to see the Semantic
Web take off in biology, then the first thing is to get as much RDF
data accessible online as we can. Waiting for the perfect ontologies
to emerge seems like a non-starter - we have to be able to handle
change as well as multiple perspectives in ontologies, so we might as
well get started with something.
2 cents
-Ben
On Jun 5, 2006, at 10:21 AM, Miller, Michael D (Rosetta) wrote:
Hi All,
But presumably the relations which characterize the structure
of UMLS could be given their own URIs, no?
Along with the concepts themselves.
And then UMLS could then be expressed in RDF, using UMLS
specific relations, rather than standard OWL relations.
This, of course, works to a certain extent but brings up an issue
that I
think will significantly impact end users and the adoption of
ontologies.
One can get a rudimentary idea of a annotation source using
standard RDF
tools through this sort of representation but that will miss the
actual
semantic value that is embedded in domain specific relations.
It would then at least be URI-ified and so connected into the
RDF-universe, and different implementors could experiment
with different approximative mappings to OWL relationships,
for some or all of UMLS, according to their particular needs.
The end user is now back to having to go to the source itself, so why
bother setting up this approximate mapping?
I've noticed a creep from several RDF representations to using these
domain specific relations which means needing a one-off parsing of
each
one to get the true semantics.
This seems to defeat the purpose.
cheers,
Michael
Michael Miller
Lead Software Developer
Rosetta Biosoftware Business Unit
www.rosettabio.com
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of
[EMAIL PROTECTED]
Sent: Monday, June 05, 2006 10:06 AM
To: [EMAIL PROTECTED]; [EMAIL PROTECTED]
Cc: public-semweb-lifesci@w3.org
Subject: RE: BioRDF [Telcon]: slides for the UMLS presentation
But presumably the relations which characterize the structure
of UMLS could be given their own URIs, no?
Along with the concepts themselves.
And then UMLS could then be expressed in RDF, using UMLS
specific relations, rather than standard OWL relations.
It would then at least be URI-ified and so connected into the
RDF-universe, and different implementors could experiment
with different approximative mappings to OWL relationships,
for some or all of UMLS, according to their particular needs.
There might, for example, be some well defined bits of UMLS
where the relations can be reasonably mapped to is_a and
part_of relations.
Matt
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] Behalf Of Olivier
Bodenreider
Sent: 05 June 2006 18:00
To: Benjamin Good
Cc: 'public-semweb-lifesci'
Subject: Re: BioRDF [Telcon]: slides for the UMLS presentation
Benjamin Good wrote:
Are there any plans to release the UMLS or parts thereof as
RDF / OWL ?
Not to my knowledge, Ben. And I certainly would be very
cautious of any
attempt to doing it. The main reason is that many relations
used for
creating hierarchies in biomedical vocabularies are not true
hierarchical relations (isa, part_of), but simply reflect
the purpose
for which these terminologies were created. For example, it
makes sense
in MeSH (i.e., for information retrieval) to have "accident
prevention"
listed as a child of "accidents". It would be wrong to assume
that all
child_of relations can be represented by subclassof
relations. And an
accurate representation of MeSH in OWL would be difficult to obtain.
-- Olivier
This email has been scanned by Postini.
For more information please visit http://www.postini.com
