POWDER tags use case rewritten and rdf:type requirement

All,

The discussion around SKOS with the SWD gave me some inspiration, and I 
found that the tags.r.us use case [1] needs some elaboration.

It is quite clear that it is hard to create something as generic as I 
need for this tagging thing, and I need to argue some more about the 
usefulness of the feature. Also, the tags.r.us use case could be a 
little more imaginatively written, so I propose the following for an 
updated working group note:

[[[
The social book-marking site tags.r.us allows their users to tag any 
resource and so provides a service through which people can annotate 
both their own and others' resources.

Anders, a zoologist and tags.r.us user, finds a website about the dahut, 
an allegedly undescribed animal that lives in the French Alps. Anders 
wants to make sure the it is understood by readers that this is a 
fictional character, but interesting to understand the full spectrum of 
cryptozoological thinking, and thus tags it "fictional".  

The word "fictional" is not very useful without context, so to enable 
such user-defined tags to be shared with others, tags.r.us allows users 
to assign a link between their own tags and a Description Resource, 
that provides the context that it is about an alleged fictional animal. 
An agent can thus use the tag as appropriate, processing the explicit 
semantics provided by the DR but perhaps presenting other users with 
Anders' original tags.
]]]

I think this should also answer Antoine's question if it is a 
important need, yes, I think it is since a tag without context is 
often totally useless to anyone other than the tagger himself. In this 
case, it would be reasonable for the tagger to tag simply "fictional", 
but that would only be useful when the context of linking that to the 
description is available.

I have another nice example of images on one of Norway's largest social 
photo sites are geotagged "drammen", which is a medium-sized city. 
There's a screenshot at [2]. It is quite clear, none of the images have 
any useful connection to Drammen whatsoever... The ability of linking a 
tag to a clearer meaning and context is important for tags to become 
useful annotation.


I have mostly looked at this as a UI problem, but it has crossed my mind 
too that the predicate could depend on the type of the object, and 
actually the subject too. It might not be quite as powerful, but then, 
exposing users to too much complexity confuses them. Antoine raised 
this point in his latest reply, and I think this creates a new 
requirement for POWDER, I suggest:

[[[
3.2.6 Distinguishable resource types

It must be possible to distinguish POWDER resource types from other 
kinds of resources.
]]]

I think this means that we cannot just use rdf:Description as we have 
previously thought, but that is a topic for POWDER to discuss, perhaps' 
in Monday's teleconf?

I hope this gives SWD a bit more to work on with respect to SKOS usage 
in tagging. I still think it is useful to have a very generic property, 
since, face it, those sites that employ tagging will not do a lot to 
create elaborate mapping systems that uses the full power of RDF. Also, 
since I have only three hours to implement it myself, I think that I 
will be stuck with skos:it. I'll document the discussion, though.



[1] http://www.w3.org/TR/powder-use-cases/#tagsrus
[2] http://dev.kjernsmo.net/tagget-drammen.png


Cheers,

Kjetil
-- 
Kjetil Kjernsmo
Semantic Web Specialist
Opera Software ASA

Received on Friday, 22 June 2007 08:57:36 UTC