RE: Minutes HCLS TC 3-22-07

Alan,

Thanks for taking the lead on this.
Wanted to add a few thoughts to the stream below..
Maybe at some point we can move over to a Wiki page.

> Qualified cardinality restrictions (QCRs) are useful all over the
> place. They allow one to classify based on there being a certain
> number of instances of a specific class as a property. The classic
> example is  that you can now say that a hand has_part 5 digits,
> has_part 4 fingers, has_part 1 thumb. Before QCRs you could only say
> hand has_part 5 things.

[VK] It will be great if you could point to the specific HC related example
here.

> Datatypes allow the specification of classes that restrict, e.g. a
> number to a certain range. So you can define "Adult" as someone whos
> age is greater  than 18 years. Useful in the clinical context, e.g.,
> for classification based on diagnostic results.

[VK] Also for consistency checking. If the normal/reference ranges for a lab
result changes, then other conditions that are based on it can be checked for
consistency

> Role inclusions allow for propagation of values across more than one
> property.  A useful example would be when using reified properties.
> This is useful in, e.g. Biopax,  where it will allow you to much more
> cleanly connect pathways to participants in the pathways. That's
> because you can express the fact that if a complex is a participant,
> and the complex has components,  then the components are also
> participants.

[VK] It will be great if you could provide a specific example.

> Punning gets rid of a big reason for using annotation properties,
> when you otherwise would
> want to use a datatype or object property with a class or property as
> subject. Punning lets you do
> that, with the caveat that, e.g, properties of classes don't have any
> effects on their instances.
> It implements this by allowing a given URI to be a name for an
> instance, a class, and a property
> all at the same time and using the usage context to decide which of
> the three is meant.

[VK] I have reservations on the use of punning, because it can be the cause of
potentially dangerous errors while designing an ontology.

On further thought, properties like curator belong to the OWL meta-model layer,
where: classes, propeties, axioms, constraints and instances are all treated as
data. I think there is a need for separation of concerns here, though I need to
do some more thinking around this to be sure.

> It happens all the same that you want to, say, restrict some property
> (say curator) to one of several values.
> Since curator makes sense for both classes and instances, previously
> you couldn't say this because
> being a property of a class meant the property had to be an
> annotation, and therefore have no axioms.
> Now it no longer does.

[VK] I would propose that it makes for good practice to define these two notions
as different properties, because the notion of curator probably has different
set of semantics when applied to class as opposed to when applied to instances.

Would be great if you could elaborate on this in the context of an HCLS use
case to better illustrate the value of doing this.

Just some thoughts -- Let us know as soon as the wiki page is set up.

Cheers and Thanks for taking the initiative.

---Vipul





The information transmitted in this electronic communication is intended only for the person or entity to whom it is addressed and may contain confidential and/or privileged material. Any review, retransmission, dissemination or other use of or taking of any action in reliance upon this information by persons or entities other than the intended recipient is prohibited. If you received this information in error, please contact the Compliance HelpLine at 800-856-1983 and properly dispose of this information.

Received on Friday, 23 March 2007 12:53:46 UTC