[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Main | Date | Thread | Author

Re: [ba-ohs-talk] Fwd: Re: [PORT-L] Digital Imagination


on 2002/03/21 7:42 AM, Jack Park at jackpark@thinkalong.com wrote:    (01)

> Context for forwarding this long post:
> I believe the following discussion relates closely to the evolving
> discussion on graph structures.  Jon Awbrey, in his usual, extremely
> articulate way, seems (to me) to be saying that ontologies, in an of
> themselves, are problematic.  But, we are still forced with the issue of
> archiving (representing) the universe for purposes of research, inference,
> and so forth.
> 
> An approach to representation being taken by the KnownSpace project
> (http://knownspace.sourceforge.net) is one that appears to be agnostic:
> simple entities, their names, values, and attributes.  Perhaps the graph
> structures we create should remain agnostic as well, and leave the
> 'interpretation' to the agents that use those structures.  I still have
> problems with that idea, and I am presently probing the limits of the
> simple entity-attribute scheme with the KnownSpace group at this time;
> possibly, my problems stem from growing up a consummate reductionist.
> 
> Jack    (02)

Perhaps it would help to clarify what is meant by "ontology". In my
experience, this term is used to refer to two very different things:    (03)

1. A knowledge or data schema used by application developers to communicate
a common set of semantics among themselves. This ontology is intended to be
frozen, and "hard-coded" into the application, i.e., it is a way to spec.
inter-operable software.    (04)

2. Information or knowledge deliberately factored out of the application
code, so that it can be maintained by a different authority than the
application developer(s), and updated independently. This is what is often
meant by "thesaurus" or "taxonomy". While such factoring is often useful, if
not critical, it does limit the extent to which the application can maintain
compatibility: if the ontology can change, then it is not a stable
"contract".    (05)

If you make this distinction, then I think most issues can be productively
re-cast as issues of which type of "ontology" a given bit of semantics
should be incorporated into...
-- 
Kevin Keck
keck@kecklabs.com
510-523-8317    (06)