W3C home > Mailing lists > Public > public-swd-wg@w3.org > April 2007

Using tags for POWDER content labels

From: Kjetil Kjernsmo <kjetilk@opera.com>
Date: Thu, 12 Apr 2007 14:07:25 +0200
To: SIOC-Dev <sioc-dev@googlegroups.com>, public-powderwg@w3.org, public-swd-wg@w3.org
Message-id: <200704121407.27408.kjetilk@opera.com>

Hi all!

Sorry for the cross-posting, but this is a topic that I think is in the 
intersection between POWDER, SIOC and SKOS, thus I would like to hear 
opinions from several groups. Reply-To is tentatively set to 
public-powderwg@w3.org, but post as you see appropriate (I guess 
anti-spam measures could make this nasty...)

I intend to support POWDER content labels on my.opera.com, since we live 
with people posting nudes and stuff. Censorship is a touchy issue. 
However, rigid taxonomies have never been popular among users, tags are 
however, so I intend that people tag their stuff, and then map that tag 
to a POWDER description. The basic infrastructure for doing this is 
allready in place: 
http://my.opera.com/semweb/blog/2007/03/08/marrying-folksonomies-and-taxonomies

However, the current interface is now rather complex:
http://my.opera.com/semweb/blog/2007/03/23/complexities-of-tag-to-vocabulary-mappin
to the extent where I think it would confuse many users and thus be of 
little value. 

So, this is how I imagine it done: A user has tags, modelled with SKOS 
concepts. Thus every tag gets a URI, 
http://my.opera.com/username/tag/nude
for example. Then, this tag needs to be bound to a resource on one hand 
and a POWDER description on the other.

sioc:topic comes to mind, thus 
<http://my.opera.com/username/albums/foo/nude.jpg> sioc:topic 
<http://my.opera.com/username/tag/nude> .

Since ICRA/FOSI will have created the vocabulary about nudity for us, 
all I want to do is to link to the description, that I hope Phil one 
day will create (this is just an example!):

<http://my.opera.com/username/tag/nude> ex:means 
<http://www.fosi.org/rdf/descriptions#just-nude> .

The first obvious thing is that I don't know what predicate that should 
be used to link the SKOS concept to the description. However, that 
could be another issue for the open SKOS issue in the core guide as 
http://www.w3.org/TR/2005/WD-swbp-skos-core-guide-20051102/#secmodellingrdf
If this approach to tag-to-URI mapping is feasible, I would be happy if 
something like that went into SKOS. The implementation I currently have 
on my.opera.com is allready a use case for this.

The next problem is that I'm not linking a resource directly to a 
description, thus making SPARQL queries and the general model more 
complex. The alternative, as I see it, is to make the UI more complex 
on the tagging services, but I don't think that's a good approach.

Thus, I think the minimum graph to link a resource to a POWDER 
description is 
<http://my.opera.com/username/albums/foo/nude.jpg> sioc:topic 
<http://my.opera.com/username/tag/nude> .
<http://my.opera.com/username/tag/nude> ex:means 
<http://www.fosi.org/rdf/descriptions#just-nude> .

and an user agent would have to understand that. I was myself hoping to 
get away with a single triple when I started to work on this, but as 
noted in my blog, it is a matter of who gets to deal with the 
complexity. So, is this the right balance?


Note that this approach does not at all use the content grouping that 
the POWDER group discusses at length, it just links resource to 
description. Of course, for large-scale workflows, content grouping is 
important, but I think "the long tail" may be more comfortable with 
this approach. It would be very interesting if we could get del.icio.us 
involved, for example.

Cheers,

Kjetil
-- 
Kjetil Kjernsmo
Semantic Web Specialist
Opera Software ASA
Received on Thursday, 12 April 2007 12:00:18 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:17:28 GMT