Re: [Bug-gnupedia] classification

[Mike Warren]

"Tom Tollenaere" <address@hidden> writes:

> Since we're discussing classifications - how about this?  Many
> scientific journals demand for 'keywords' to help readers figure out
> which article could be interesting for them.  How about foreseeing
> something similar (should be easy adding the tags in the XML)?
> 
> Trouble is that people tend to 'invent' their keywords, so should be
> have a fixed list of acceptable keywords?  I'm not sure about this,
> as having a list also implies maintaining it...  Any thoughts?

I liked Bod Dodd's suggestion of keeping the classification separate
from the actual article content. One might consider making a
``classification layer'' (which could very well have its own large
database associated to it) which would put the actual articles into
classifications.  Multiple classification layers are then possible, to
suit different needs (i.e. Dewey Decimal and Library of Congress have
already been mentioned) with no trouble whatsoever.

All the actual content needs is some unique ID so that the classifiers
might refer to them. This also raises the possibility of a
``reader-based'' classifier where the readers could vote on better
classifications for the content. Also, a group of ``expert'' editors
could get together and make a classifier. This allows all the
advantages of editing (less content which is irrelevant) and none of
the disadvantages (an appearance of ``censorship'' to some or the very
real possibility of bias). If some different group doesn't like the
classifications of that group, no problem: they can make their own!

Classifiers could obviously ``censor'' information by classifying it
to a ``garbage'' or non-visible category, but this isn't a problem, as
the real, actual article could remain in the database (and be visible
under a different classifier). I still think removing some articles
might be necessary, though (perhaps: if it hasn't been viewed by any
classifier in over a year, delete it?).

After all that though, keywords in the XML might be useful for the
classifiers.  I wouldn't suggest limiting these to some fixed list,
though. If an author wants to invent some key-words for his article,
where's the harm? Indeed, what if the ``fixed list'' which gets chosen
is a bad one?

A particular classifier could obviously limit itself to only look at
certain keywords, if this was desired...

-- 
address@hidden
<URL:http://www.mike-warren.com>
GPG: 0x579911BD :: 87F2 4D98 BDB0 0E90 EE2A  0CF9 1087 0884 5799 11BD