[Sigia-l] Eliminating categories in favour of tagging

[Alexander Johannesen]

On 3/10/06, Dmitry Nekrasovski <mail.dmitry at gmail.com> wrote:
> My employer is in the process of upgrading its internal blogging
> platform. The new version announcement came out a couple of days ago,
> and contained a statement to the effect of "since tagging is the new
> best practice in metadata generation for blogs, the new version will
> not have post categories, only tags".
>
> What does everyone think of this move from an IA standpoint? Please
> don't get caught up on the "best practice" part... :)

Hmm, who was it that coined "tagging is poor-mans uncontrolled
vocabulary"? Maybe I just did, but it's a good quote. :)

Well, apart from being a best-practice thing, it's also a misguided
one. First of all, a category system is an attempt to unify
categorisation, while tagging does the opposite. A category system
ensures that everything must fits into those categories even when they
probably shouldn't (a problem we're trying to solve), while tagging
creates one category for every plural and spelingmisstake there is (a
problem with the problem-solver). In essense, tagging is an attempt at
organic faceted organisation, and sure it works for simple things but
as soon as you *need* those tags to have applicability within
controlled systems, you need to have some form of concesus on them.

Recent thinking (actually, if you move in library circles it ain't
recent at all, but hey ...) involves 40% pre-defined categories (2 to
3 levels, depending on how brave you are) and 60% tagged
sub-categories, with various degrees of flexibility within those
percentages. It's based on the thinking that if a tag is mismatched it
at least has a few parents that will be valuable and workable, so
tagging within a simpler category system. Another way to do this is to
make tags a part of a dynamic controlled vocabulary, but this requires
some technical implementation thinking as well as a purely knowledge
management thinking.

Of course, blind tagging (non-classified) is a basic faulty faceted
system and works for hobbyists and with small datasets, so if your
stuff doesn't need that degree of absoluteness or is reasonably small
and not too diverse in scope, I don't see any problem with it.


Regards,

Alex
--
"Ultimately, all things are known because you want to believe you know."
                                                         - Frank Herbert
__ http://shelter.nu/ __________________________________________________