Re: SEM: comprehensive entailments without dark triples

Dan Connolly wrote:
>
> So the cyclic structure above isn't entailed, under
> Jeremy's axioms. So what? What do I, as a real-world
> user, lose?

How about, from my perspective as a real-world user, working in a real-world
problem domain that currently uses description logic based classifications
(e.g. SNOMED-RT, etc).

Suppose there exists a disease which is defined as a cyclic structure.

Suppose I have an individual defined by a set of facts, and these facts
support the classification of the individual as being in the class of
patients having the disease.

It would be really really great if, after going to all this work to develop
this disease classification system, the classifier were able to tell me that
a patient having all the pathognomonic symptoms of a disease actually has
this diagnosis.

So what I would stand to lose is that what we do would not be applicable to
the 200,000 - 400,000 term ontologies created and in clinical use -- not to
say the least making inferences _across_ different healthcare ontologies
e.g. MeSH, CPT, ICD, SNOMED, GALEN etc.

(Figuring this out in an iterative fashion would take something approaching
infinite time).

Jonathan

Received on Monday, 22 April 2002 15:15:46 UTC