[topicmapmail] Superclass-subclass indentation in the Omnigat or

[Robert Barta]

On Thu, Jan 09, 2003 at 01:55:23PM -0500, Dichev, Christo wrote:
> RE: instance-of, class-subclass discussion again:
> 
> Composite objects are characterized by the structure of their 
> components, that is, by the components and how those components 
> are related. Normally we use "part-of" (a transitive and reflexive) 
> relation to say that one component is part of another.

Dichev (or is Christo your first name?),

I am not perfectly sure whether 'is-part-of' should be reflexive,
but...

> My question 
> is how "part-of" relation can be expressed in terms of TM i.e. 
> "Reinforcement learning" is part of "Machine learning"? I think 
> that this type of "structural relations" can not be expressed in 
> terms of  "class-subclass" or "class-instance" relations. Part-of  
> plays important role in ontology - "class-subclass"  deals with 
> categorization, while part-of deals with the object-components topology.

..., yes, for 'is-part-of' and quite a few more you will need to
introduce a dedicated association type.

I think the standards people did a _good_ job not to hard-code a set
of relations in the XTM standard, like

  - component-member (wing/airplane)
  - member-collection (tree/forest)
  - portion-mass (slice)
  - place-area (city/country)
  ....

They seem to be far too application specific. Even 'is-part-of' is
different for

  #  "Reinforcement learning" is part of "Machine learning"
  (is-part-of-1)
  part  : reinforcement-learning
  whole : machine-learning

and

  # parketry floor is part of my house
  (is-part-of-2)
  part  : parketry-floor
  whole : rho-house

In every-day conversations we are not distinguishing this because the
other human (if there is one) is disambiguating. But for an
association the context is very limited. Here the TM author has to be
a bit more disciplined.

\rho