[TagCommons-WG] mechanisms for sharing tag data

Nitin Borwankar nitin at borwankar.com
Fri Mar 2 08:46:57 PST 2007


Richard Newman wrote:

>
> On  1 Mar 2007, at 5:31 PM, Nitin Borwankar wrote:
>
>> I am wary of acccepting a single database 
>
>
> Nitin,
>   I think you're misunderstanding the word 'ontology' in this  
> context. An ontology describes the conceptual model of a domain, a  
> model that an interface programmatically encodes. The idea of an  
> interface that doesn't commit to a single ontological model is  
> nonsensical.
>
>   Of course, the choice of URIs in a particular RDF encoding of such  
> an ontology is a different matter, but the ontology itself  
> necessitates quite a firm contractual commitment from a participant.  
> For example, that a tagging event is some kind of relation between a  
> tagging party, a tagged resource, and a tag is a core part of such an  
> ontology.
>
OK, agreed that my previous statement does misunderstand what is meant 
by 'ontology'.

To clarify, I assume  that we all have a common subset of things, types 
and relationships in our tagging models,  and that each of our models 
diverges from this common subset in our specific domains.  Implicit in 
my statements is the assumption that we express these relationships in 
common English language first in our glossary.
This is the intersection model if we can arrive at it.

Call that an informal core ontology if you want.  The initial interfaces 
would represents operations on the ontology expressed in the glossary.
At this level I don't understand the added value of using the word 
ontology rather than using language used in creating a data model.
Formally I understand that even at this stage we must call it an ontology.

My understanding (or misunderstanding) of the use of ontology beyond 
ordinary data modeling is that in specific application domains,
more domain-specific  relationships are added to the core relationships 
and relationships more complex and expressive than has-one, has-many are 
constructed to properly
cover the domain. If we want to call the initial common data model an 
ontology that's fine - but I was using the word where I see 'ontology' 
adding more than just the
has-one, has-many relationships of simple data models.  Flickr's 
ontology would add things, types and relationships  unique to photos but 
not useful in discussion of bookmarks.   If this understanding of 
ontology is wrong  then  I need to 'go look  it up'.

So to restate - we should arrive at a simple core ontology that we can 
express in a) english in the form of a glossary and b) a data model. ( 
ER diagram, RDF ...)
These are required for interoperability between public web applications 
currently practising tagging in specific application domains.
The intent is that the core will be useful to all or most of the 
existing apps.

Then we add interfaces that allow common CRUD operations on this data model.
I keep reverting to describing it as a data model because the way I was 
looking at the core - it would contain relationships expressed as 
has-one and has-many only.

Sorry for abusing and loosely using the term ontology,  I get pretty 
upset when people misuse formal terms in my domain as well.

Nitin.

>
> I think that the difference in our viewpoints comes from a difference  
> in our definitions of "ontology".
>
Yes, and I hope there's less of a difference now.

-- 
Nitin Borwankar

http://greener.com
Find, Learn, Act .... Greener, the search engine for the planet

http://tagschema.com
Implementation of tag database applications

nitin at borwankar.com
510-872-7066




More information about the Wg mailing list