[mb-users] Brainztorming: rank edits, not users

[Aaron Cooper]

On Wed, Apr 30, 2008 at 10:38 AM, Philip Jägenstedt <philip at foolip.org> wrote:
> Hi all,
>
>  I had an idea while showering this morning. There's been some talking
>  about adding new user levels between the standard editor and
>  autoeditors, or otherwise provde a ranking system to tell the bad
>  edits from the good.
>
>  Objective: find those edits most in need of checking (separate noise
>  and signal in the edit queue)
>
>  I'm thinking that instead of ranking users and changing the voting
>  periods, we could rank each edit individually based on a numer of
>  factors, where the user is only one.
>
>  Examples:
>
>  Edit starts at rank 0
>
>  - if the edit is from an editor with less than 10 votes, increase rank
>  dramatically
>  - if the edit is from an editor with high failed edit ration, increase rank
>  - if an has no-votes but no comments from no-voters, increase rank dramatically
>  - if add release edit is missing release event, increase rank
>  - if add release edit is missing discid, increase rank
>  - if add release... (you can think of many parameters here)
>  - if add wikipedia AR links to article of very different name, increase rank
>  - if removing release with discid, increase rank
>  - if capitalization doesn't match what guess case would give, increase rank
>  - if script doesn't match with actual content, increase rank (e.g.
>  check if all characters are in unicode latin blocks)
>
>  I think you can imagine many "filters" which can be applied
>  automatically and give an indication of the quality of the edit.
>
>  WDYAT?
>
>  Philip


I like this idea a lot.

This would allow us to focus our attention on edits which most likely
need the attention.  Other things that should increase the rank of the
edits are:

* Language or Script unset in AddRelease edits.
* Perhaps a slight increase for any AddRelease edits without DiscId
and without edit note from editor.

-Aaron