Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: SUO: Monolithic ontologies (was ontology as science)




Pat,
   If you had an application that required you to reason about a muscle as 
both a BodyPart and Food you could simple define a particular instance as 
such.  This might be unsatisfying to someone who wants (for some reason) a 
pure treatment that somehow separates out classes from contexts and roles, 
but I don't see that it causes a practical problem in knowledge 
representation.  Do you have an example that shows a problem?
   I don't think the issue of roles is particular to a 3D approach, but to 
your specific question of how to represent PartTimeStudent (leaving aside 
for the moment whether that deserves to be reified), it would presumably be 
a &%subAttribute of &%Student.
   To specify the characteristics of Student, one could use axioms, just as 
for the definition of any other term, for example

(=>
   (attribute ?X Student)
   (capability EducationalProcess patient ?X))

Adam

At 10:49 PM 7/1/2003 -0400, Patrick Cassidy wrote:

>Adam,
>   Thanks for the response.  A follow-up here:
>
>Adam Pease wrote:
>[Pat Cassidy wrote]
>
>>>    If roles cannot be classes, then do you consider SUMO
>>>classes like "Food" and "BiologicallyActiveSubstance"
>>>as *not* being roles which might change for a particular object?
>>
>>I'm guessing that you're concerned about cases like the same substance 
>>being used as a floor wax or a dessert topping, but as I see it, it would 
>>still be a food, just as a table could be used as a chair, or a TV as a 
>>hammer, but they would still be tables and TVs, respectively.
>     I'm not particularly concerned about artifacts, it is just
>certain things like "Food" that I wonder about.  The gloss for
>"Food" in SUMO reads:
>
>  "Any &%SelfConnectedObject containing &%Nutrients, such as 
> carbohydrates, proteins, and fats, that can be ingested by a living 
> &%Animal and metabolized into energy and body tissue."
>
>    The muscle of any live animal would seem to qualify, including
>human muscle.  Wood is food for termites.  It would seem that
>any part of any living thing can serve as food for some
>animal, so if Food is not a role that depends on context,
>what is the distinction between Food and living things generally?
>Could you not just have "Organism" as one subclass of "Food"?
>(Even bacteria can be food for protozoans, and fungi for ants).
>
>
>Another class that to me has some character of a "role" is 
>"BiologicallyActiveSubstance".  The gloss reads:
>
>"A &%Substance that is capable of inducing a change in the structure or 
>functioning of an &%Organism.  This &%Class includes &%Substances used in 
>the treatment, diagnosis, prevention or analysis of normal and abnormal 
>body function.  This &%Class also includes &%Substances that occur 
>naturally in the body and are administered therapeutically. Finally, 
>&%BiologicallyActiveSubstance includes &%Nutrients, most drugs of abuse, 
>and agents that require special handling because of their toxicity."
>
>     Consider a synthetic substance invented by a chemist, which years
>later is discovered to have therapeutic activity. Was this always a 
>"biologically active substance"?  How does one classify substances
>used traditionally in medicine which are discovered to actually
>have no biological activity at all?  If a species evolves to
>become sensitive to something that previously had no effect
>on organisms (say, an allergen), would that also always have
>been a biologically active substance?
>     I suppose you can use the "uses" relation in specific instances
>for some of these cases that are borderline, but without a class
>that expresses a role, it's not obvious how all these things can be
>related to each other.
>
>
>As to roles of humans:
>
> > Those sorts of roles have been treated as properties.  Take a look
> > at SocialRole
><http://ontology.teknowledge.com:8080/rsigma/SKB.jsp?req=SC&name=SocialRole&skb=SUMO>
>
>   OK, so using this it would seem that "Student" would have to be an
>instance of a SocialRole.  Is that right?  If "Student" is not a
>class, how would  SUMO express the relation between "Student" and
>"PartTimeStudent"? "collegeStudent"?  Or between "Physician" and
>"Surgeon"?  How do you specify what the characteristics of
>a "Student" are?
>
>    Most ontologies I have seen allow many roles as classes.  I'm
>curious how one can do without them in 3D.
>
>     Pat
>
>=============================================
>Patrick Cassidy
>
>MICRA, Inc.                      || (908) 561-3416
>735 Belvidere Ave.               || (908) 668-5252 (if no answer)
>Plainfield, NJ 07062-2054        || (908) 668-5904 (fax)
>
>internet:   cassidy@micra.com
>=============================================