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

Re: Tag ontology RFC

From: Richard Newman <r.newman@reading.ac.uk>
Date: Tue, 5 Apr 2005 10:56:03 +0100
Message-Id: <db256524884d446d4fc041bdcb630ef4@reading.ac.uk>
Cc: semantic-web@w3.org
To: Danny Ayers <danny.ayers@gmail.com>

Yes, I'd considered that possibility --- having a (possibly shared) tag 
applied in a reification is more or less the same as each person having 
their own tag and using a direct triple*. There are a few minor 
differences (I can deliberately apply someone else's tag; open world 
assumption and all that -- more in the footnote), but the main one is 
the "whose tagging is this?", which is dealt with by the namespace.

I thought it safest to split them apart --- it stops me accidentally 
saying

chimp:bubbles tags:tag danny:monkey .

and asserting that you know something about Bubbles, when I really mean 
that I associate Bubbles with a monkey tag.

It's still possible, of course, to use personal tags with the reified 
approach -- it just uses a few more triples than the non-reified way :D

I quite agree with your last paragraph. You sum up quite well the 
crisis that led to me leaving in the simplistic <tags:taggedWithTag>, 
which is the "straight tag" approach, but I'm quite conscious of 
doubling the effort of tool producers. Thus, I am torn. :)

Rules will sort it all out.

?x tags:tag ?y
?y tags:associatedTag ?z
?y tags:taggedBy rich:RichardNewman
=>
?x tags:taggedWithTag rich:(uri_tail ?z)

Thanks Danny,
-R

* If (a) there is a direct mapping between rich:chimp and danny:chimp 
(probably based on literals),
      (b) "rich" can't tag things with danny:chimp (without being aware 
of the consequences), and
      (c) we use other measures to associate e.g. dates with the tag, 
then they are perfectly isomorphic.

On Apr 5, 2005, at 10:42, Danny Ayers wrote:

> On Apr 5, 2005 10:53 AM, Richard Newman <r.newman@reading.ac.uk> wrote:
>>
>> All,
>>    I threw some rough notes about the tag ontology on the Web for 
>> future
>> reference. I might yet also get this thing out of the door, but I have
>> a paper to write that I've been putting off :)
>>
>>    <http://www.holygoat.co.uk/projects/tags/>
>>
>>    Further comments always welcome.
>
> Cool, nice docs. Heh, now this /doesn't/ correspond directly with the
> way I'd have modelled it, but I won't hold that against you - I think
> it's more or less isomorphic, or at least interoperable.
>
> Where you have "universal" tags, tag:chimpanzee, tag:monkey, all in
> the same namespace,  I'd have opted for per-creator tags,
> richard:chimpanzee, danny:monkey etc.
>
> Ok, in the universal approach, presumably the Tag instance is tightly
> bound to the string - tag:chimpanzee always relates to the literal
> "chimpanzee". In the personal approach, the Tag instance is bound to
> the concept a specific person associates with the literal
> "chimpanzee".
>
> But I reckon there is like to be some level of equivalence between say:
>
> http://example.org/item123 :tag [ a :Tagging ;
>    :associatedTag tag:blog, tag:chimpanzee ;
>    :taggedBy <http://example.com/People/Jim> ;
> ] .
>
> and
>
> http://example.org/item123 :tag Jim:chimpanzee .
>
> It's a funny sort of a situation really - I could certainly imagine
> the more granular approach being better for general inference (you
> mention the author's opinion), but the direct version might be more
> useful for (quick & dirty) queries. But the nice thing is that I
> reckon that both approaches could co-exist, class equivalences/subs
> could be set up between say tag:chimpanzee and Jim:chimpanzee. There
> is a big difference between using Jim:chimpanzee and stating that Jim
> did the tagging, but I'm not sure that would be a problem in practice.
>
> Cheers,
> Danny.
>
> -- 
>
> http://dannyayers.com
Received on Tuesday, 5 April 2005 09:56:10 UTC

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