W3C home > Mailing lists > Public > semantic-web@w3.org > March 2005

Re: Tag ontology RFC

From: Richard Newman <r.newman@reading.ac.uk>
Date: Fri, 25 Mar 2005 19:22:27 +0000
Message-Id: <631f5bc99f5403b556408eeb4ab07a30@reading.ac.uk>
Cc: semantic-web@w3.org
To: Seth Russell <russell.seth@gmail.com>

> I would say that Tagging is a sub class of Items.

Quite possibly; there's clearly some overlap. Subclassing is a lot 
safer than direct reuse, which is what I got out of your original 
message: "After working with this i am pretty convinced that a Tagging 
is a RSS item".

> I think we need to ask how is this ontology to be used?  What is it's
> practical value?  What is going to prompt people to write meta data in
> this ontology?  But we cannot ignore what is already happening with
> folksonomy.  The feed over at del.icio.us is already about 10 taggings
> *per second* - to verify that just go over to http://del.icio.us/ and
> keep reloading the page.  Now consider all the other
> folksonomy.servers and you already have a Mississippi river of
> taggings flowing.  Each of these taggings is collecting the meta data
> that your ontology is addressing.  Now are you going to just ignore
> that gigantic river and make up some semantically pure ontology?

Absolutely not, which is why I went the public route rather than just 
hacking something together for my own software. I see interoperability 
as the key aspect of the whole SW, and it's one of the core parts of my 
research.

However, I'm a cautious ontology designer by nature; I recoil in horror 
on seeing transitive and symmetric properties, for example, when they 
fail to model the domain accurately; you ruin the utility of logical 
inference. That's my worry with Item, and with other direct reuses. In 
general, it's safest to design as Danny suggested -- to model the 
domain correctly, and then worry about interoperable mappings to other 
ontologies.

There's a real `danger' that there's an ontological mismatch between 
tags and items, and this will somehow blur the lines between RSS feeds 
and tags _in an undesirable way_ -- it needs some attention*. The fact 
that del.icio.us generates RSS feeds (that happen to be RDF) doesn't 
_necessarily_ mean that this serialisation into RDF is `correct'; it 
just so happens that it needs to be that way to be read by RSS readers.

I think there are important relationships between Annotea's work, 
bookmarks, and, indeed, RSS1.0 (and RSS1.1, presumably). But I want to 
thoroughly study the specs for each before I write "tags 0.2". In 
summary, interoperability is vitally important -- but it's pointless if 
the output is wrong.

Thanks,
-R

* this shouldn't happen with subclassing, but you can imagine the 
situation with direct reuse of Item -- every RSS feed you read would 
dump its items into your store's tags, which would really clutter an 
interface, and presumably not match your intent...
Received on Friday, 25 March 2005 19:23:06 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 07:41:45 UTC