[Commons-l] Complete brokenness of categories is making me mad.

[Magnus Manske]

On 8/14/07, Gregory Maxwell <gmaxwell at gmail.com> wrote:
> In short, our category system is completely screwed up.
>
> I've complained about this in the past, but it only seems to be getting worse.

I don't think (for once) that this can be solved by throwing code at
it. As Brianna has written time and again, we need something that is
both category tree and tag cloud, somehow.

Qualities such a system need would be IMHO:
* easy assignment by users
* ability to be queried for general properties ("flower")
* ability to be queried for specific subsections ("yellow rose")

So, here is my suggestion of the day:
* users can assign tags (single words or phrases) to an image
* users can create implications of tags
* implications are fulfilled by the software automatically
Example: I add the tag "MIG-29" to an image. Someone has said (or will
say) that "MIG-29" implies "military aircraft", "russian aircraft",
"supersonic aircraft". These, in turn, imply "aircraft" (each of
them). All these tags will be added to the tag list of this image
automatically (not in real time, but through an updating process in
the background). Likewise, if I add an implication to a tag, all
images carrying this tag will be updated automatically.

This would effectively push the "category flattening" from serachtime
to creation time. It will also generate a s**tload of tags for each
image, but "implied" tags could be hidden from view (though not from
search) by default. So, your image would be part of queries for
"aircraft", "military aircraft", or "russian aircraft". The more
parameters you add, the mode precise the query gets. The intersection
could probably be done by your tool, with minor adaptations, very
quickly as well.

We could even use the existing category system to fill the
"implication rules", so all that work was not in vain.

Magnus