On 4/22/11 4:40 PM, Daniel Schwabe wrote:
Dear all,
I'd like to take this opportunity to change the focus of the discussion.
It seems to me this discussion is addressing a recurring problem we have been
facing in the Semantic Web and, more pragmatically, in Linked Data - the way
the ontology is specified and the way the ontology is*communicated* (both
vocabulary and instances). The various arguments pro/against each of the
options posed in this thread are a good example that it is too difficult to do
both at the same time.
The ontology specification can have more rigid, perhaps less human-friendly
syntax, as it should be machine-processable. It makes sense to keep it as
concise and simple as possible.
On the other hand, communicating about it, in various contexts, is a much more
complex endeavor, and I believe it is unlikely that a single solution/approach
will suffice for all situations.
Therefore, I propose that we should*separate* vocabularies dedicated to specifying how an author wants
to*communicate* (i.e., "talk", "present", "discuss", etc...) about an
ontology. It should have a way to, among other things, better specify the particular types of context this
form of communication for which this form is intended, and one could have many of them. In a way, something
analogous to media types in CSS, but much more sophisticated. These presentations should not add any formal
semantics to the resources described (yes, I realize this can be debatable from a more philosophical
standpoint, since any form of communication necessarily adds some semantics - added by the recipient/reader,
which may not have been exactly what was meant by the author )
Then one could attach a property (e.g., iv:description) to a resource (including e.g.,
terms in vocabulary specifications) which would specify how this resource is to be
"communicated" (e.g., displayed, spoken, etc...).
If this becomes standard practice (e.g. how rdfs:label is today), tool makers can take
advantage to produce better designer-oriented "pretty prints" of an ontology,
or end-user-oriented versions, etc..., without changing the ontology definition itself.
Thus, human readability of a vocabulary specification would not be such a major
requirement as it seems to be, currently - Most people, including developers,
would work with them through some specialized presentation form, most
appropriate to his/her context.
So perhaps we should start focusing on what these "communication" vocabularies
would look like...
Just my 2c
Daniel
All,
Like Data Quality, this is yet another important topic of discussion.
Daniel: I've used you post to open up this very important thread of
discussion.
--
Regards,
Kingsley Idehen
President& CEO
OpenLink Software
Web: http://www.openlinksw.com
Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca: kidehen
