How many versions of an ontology are allowed? [Was: Weakness in the Semantic Web?]

Hi Folks,

Thanks for all the responses to my post regarding the allowed amount of
diversity in ontologies.  I have assimilated all your comments, and
summarized them below.  Please let me know if I have accurately
summarized your comments.

Issue: How many versions of an ontology are allowed?

Example: How many versions of a Camera ontology are allowed on the Web?

There are two ends of the spectrum:

1. Only a single Camera ontology is allowed - there is universal
agreement on a Camera ontology.

   Comment: in a highly distributed, autonomous environment such
   as the Web, this is not realistic.

2. Every application has its own Camera ontology
(ontology-per-application)

   Comment: zero interoperability!

Here is the middle ground:

The number of Camera ontologies is greater than one, but less than
one-per-application.

The Evolution of a Web of Camera Ontologies

Consider this scenario: A community comes together to create a Camera
ontology.  Within that community there is perfect interoperability.
  
Independently, another community comes together to create another
version of the Camera ontology.  Within that community there is perfect
interoperability, but across the two communities there is zero
interoperability.

Someone in the first community recognizes that a class in that
community's Camera ontology is equivalent to a class in the other
community's Camera ontology, so this equivalence is added to both
ontologies (using owl:equivalentClass).  Alas, a bridge has been created
between the two Camera ontologies, and now there can be interoperability
between the two communities.

Ultimately, all the different Camera ontologies get linked together, and
there is global interoperability.  TaDa!

Now going the other direction (from macro to micro) ...

Within a community there may be subcommunities which refine the
community's Camera ontology.  Provided the subcommunity's ontology is
linked to the community's ontology then there will be perfect
interoperability.

Thus, there can be diversity in ontologies both at a micro level as well
as at a macro level.

....

Well, that's my summary.  What do you think?  /Roger

Received on Monday, 5 May 2003 09:28:00 UTC