Open GoogleCodeExporter opened 8 years ago
There is indeed a general issue in OGMS concerning the connections between some
disorders (like "inflamed blood vessel") and the corresponding anatomical part
("blood vessel"), which comes from the fact that disorders must be independent
continuants (if I remember correctly, Ludger Jansen suggested that disorders
could instead be qualities). Here, this is a case where it would seem simpler
to have a quality (inflamed) as a disorder, but that’s not OGMS approach.
As you noticed, we chose most of the time a very general relation (more general
than e.g. a parthood or is_a relation) between a disorder and its corresponding
anatomical part, namely located_in. This way, even if we change the label of
the disorder to something else, this relation would probably continue to hold.
More specific relations (like parthood or is_a) can be added in the future when
applicable.
We also wanted to respect the principle of single asserted inheritance – this
is why we did not assert "inflamed blood vessel" as a sub-class of both
"disorder" and "blood vessel". We could try to make such that one of these
inheritances would be inferred, but this would require a heavier axiomatization
than the one we have developed yet.
As you suggested, "blood vessel inflammation disorder" would fit better the
role of disorder (by definition), but would be less informative, so on the long
run I am not sure that this would be the best choice. Our choice of "inflamed
blood vessel" should not be understood as a definitive position on this
problem, but as a suggestion that could be improved and that raises questions
about OGMS approach.
Original comment by adrien.b...@gmail.com
on 15 Mar 2013 at 6:05
- I'm not an expert on OGMS, I find the treatment of disorders as ICs
confusing. In this case I'm not sure if in OGMS the disorder is part of the
blood vessel (which part? How do we draw the boundaries?) or the blood vessel
itself (which is different than how you have modeled it).
- it isn't true that located in is more general than is_a (it isn't really a
well-formed statement in OWL, as SubClassOf is not an object property).
- if you do decide to model this such that "inflamed blood vessel" is a subtype
of "blood vessel" then you do not necessarily have asserted MI. You model this
as equivalentTo "blood vessel and has_quality some inflamed" and then you get
inferred parentage to "inflamed vessel", "inflamed organ", "vessel with
abnormality" etc. I'm not saying you should model it this way for now - if
you're using OGMS you should use it in toto - ask the OGMS experts for help
here.
- The no asserted MI thing has become dogma. IMHO the important thing is
understanding how to use reasoners to do the classification work for you, and
having you avoid recapitulating the work of anatomy ontology. See issue #1
- I see your point about the naming, my suggestion "blood vessel inflammation
disorder" is horrible. But I think it is confusing how it is labeled right now.
The label suggests one thing, the OWL another. Perhaps it could just be marked
with a comment in the interim.
Original comment by cmung...@gmail.com
on 15 Mar 2013 at 6:25
Original issue reported on code.google.com by
cmung...@gmail.com
on 15 Mar 2013 at 12:29