Re: Is ontology an information resource?

On 2007-10 -21, at 11:50, Alan Ruttenberg wrote:

> On Oct 21, 2007, at 11:11 AM, Tim Berners-Lee wrote:
>
>> But you should hash URIs for normal ontologies.  They are simpler  
>> and faster at run time, and are served up just be writing a file  
>> and serving it up.
>
> Hi Tim,
>
> What's a normal ontology?
>

An ontology of a finite number terms agreed by a community to have a  
common;y understood meaning, to be used in communications between  
programs used by the community.

Examples:  RDF, RDFS, OWL, Calendar, FOAF, and ontologies  
corresponding to things like iCalendar, vCard, OFX, XBRL, etc.     
With these things the value is in the instances, not the ontology --  
the data is being generated all the time, but the ontology changes  
very rarely.

(Counter-example:   Snomed.  Here a very large number of things  
(anatomical parts, diseases, etc) are related.  The collection does  
not so much provide a language cfor communicating as the actually  
knowledge of the community. The value  is in e ontology itself, which  
changes  frequently as community knowledge is revised.)


> In my view this choice makes it easier for the provider, and harder  
> for the user, since, as I've mentioned, it means that a user who is  
> interested in just the term has to fish through a file and figure  
> out what's relevant for understanding that term from among the rest.

Well, as someone who  considers myself an ontology user I must say the

Have you programmed using an RDF API?
There really is no 'fishing around'. RDF systems parse the ontology  
into a graph, and index it by the terns, so you can instantly see all  
the things the ontology says about a term.

A lot of RDF users also happily look at the N3 of a graph say also to  
read the ontology directly.  When have you found yourself doing a lot  
of 'fishing'?


> In general, I think we should be encouraging practices that are  
> oriented towards making it easier for the user, even if that takes  
> a little more effort on the part of the provider.
>

When I use ontologies, I mostly do it through the software of  
course.  So for example the tabulator uses ontologies to present data  
to me.  All that I need is for the necessary data to be pricked up  
promptly and efficiently.   In this case 303's for each term (like  
for Dublin core and FOAF) take an order of magnitude more network  
round trips to process compared with hash ontologies (like RDF, RDFS,  
OWL, contact, gen, etc)

Hash ontologies users can read, process and cache.  For example, many  
hash ontologies I have on my laptop, and can use these persistent  
caches when I am on a plane.    The web technology around me isn't  
set up to cache 303s.

Tim

> -Alan

Received on Sunday, 21 October 2007 18:44:32 UTC