I will describe my idea of creating categories through voting in points (still a kind of draft): // implementation notes are marked as here
Vote for category link on every page can lead to Add keywords to the article you edit // only one extra box needed
Software will count keywords as votes for particular categories. // only one table for categories is needed // cat_id, cat_name, cat_votes // keywords can have mulptiple words like "physical anthropology" // so some separator is needed
Keywords added by person who done some edit are weighted higher e.g. like 5 'viewers votes'. // this extra box can appear under editing pane like // the description box appears
It is not obligatory to add any keyword in this case.
At the top of the article all categories can be presented in the order of descending number of votes (let's say top 10).
If you disagree with the fact of a given article being in some category or so high in the hierarchy preceed a keyword with minus sign.
Everybody (and especially admins) can view the whole list of categories that a given article was assigned to (the number of votes are presented as well). // maybe Special:Assignedto page?
Admins can remove real rubbish links to categories (including vandalism). // delete link on said Special:Assignedto page (or a series of check-boxes)
Removed links are logged on a special page - that way everyone can check if some admin overuses his rights.
=== Sifter extension ===
Another voting system could be implemented for "certified" users who can vote for quality. This could incorporate "lighter" idea of sifter into Wikipedia.
This won't have influence on editing capabilities of anyone.
Sister voting system, third one, could be implemented for "certified" users who will vote for their category scheme.
Everyone can select either first or the second category scheme or both when viewing article. As well, anyone can select quality note to be rendered on the page.
----
Open problems: - Will people like it? - How filter RecentChanges with so many categories? - What if a vandal, insane user, etc edits just to promote an article in a given category? How to decrease voting counters? - Should we stick to stif tree-like structures? (Inclusion of one category into another can be done through... voting on Category:Foobar_category page that describes this particular category)
Regards, Youandme
---------------------------------------------------------- Think fast! This message will self-forget in 15 minutes ;)
I've been thinking a lot about rigid categories (clear-cut, well defined rules, often mutually exclsuve and/or hierarchical) and soft categories (fuzzy membership rules, rarely mutually exclusive or hierarchical).
It seems to me that we want not categories, but ''relations'' between objects, from which categories fall out as a consequence.
* Relations should be of the form x R y, where x and y are articles, and R is a relation-name. I propose a new namespace, Relation: for talking about relations. Notice that this makes it possible that x or y, or both, can also be relations!
The current link structure defines exactly one kind of relation: links-to
* Relations need not be equivalence relations, or even be reflexive or transitive.
* Now reserve a special syntax for putting relations into articles:
#relation [[article 1]] relation-name [[article 2]]
Note that article 1 or article 2 should be the name of the article that the line is written in: perhaps we could have a shorthand, 'this' for this article.
Now we can say things like:
#relation [[science]] is-a [[field of enquiry]]
#relation [[physics]] is-a [[science]]
#relation [[England]] subset-of [[Great Britain]]
#relation [[United Kingdom]] is-a [[nation state]]
#relation [[two]] is-a [[prime number]]
and so on...
It's ideal fodder for machine analysis (RDF can be generated naturally), and easy to parse and write. It's easy! It's wiki-style!
* Now comes the nice bit.
We can have meta-rules:
#relation [[relation:subset-of]] reverse-of [[relation:superset-of]]
#relation [[relation:subset-of]] more-important-for-sorting-than [[relation:member-of]]
#relation [[relation:identical-to]] is-a [[equivalence relation]]
Brainstorming...
* fuzzy relations:
#relation [[man]] indentical-to [[woman]] 0.85
* Bayesian inference on fuzzy rules...
-- Neil
wikipedia-l@lists.wikimedia.org