Hey all,
my opinion, based on the 7000 images last year in Belgium.
1/ some preprocessing. Remove images that are too small, have no emailaddress linked to their account, they cannot participate, ...
2/ User generated voting. Two types of votes:
- *failed images* (we need to define and explain this very well!) Three "-1" mean the image is removed from the contest. This cleans up the list of images, so people voting at the end will see less images then the first ones. This greatly improves the voting. - *favourite images of the voter*. Every selected image is giving 1/N points (N being the number of votes giving by this users*). People who select a limited amount of images will impact one image more then somebody who selects almost all images.
3/ The final selection happens by the judges (you cannot replace this by user voting). They only see the best 100-200 images.
Regards, Maarten
* for the mathematicians, I would keep 10 < N < 100, to avoid strange results when people only select one image, and to give more credit to people selecting more than 100 images.
2012/8/7 Lodewijk lodewijk@effeietsanders.org
Yeah, it won't be easy :) But with big numbers it might be the only way. One possible way is to only allow two scores: 1 or 0 (and skip). And then take the average instead of the sum. That way you should be able to get the most wonderful images on top - we don't need to rank everything, but should only make sure that the top-25 images are within the selection of 500.
Lodewijk
2012/8/7 Platonides platonides@gmail.com
On Tue, Aug 7, 2012 at 1:01 PM, Lodewijk lodewijk@effeietsanders.org wrote:
My idea of making the number of jury members in a particular round
infinite,
was that we could hand out a jury token to each community member
easily, and
let them process 100-1000 images. If enough people do that, and if the distribution is either random or based on the pictures that have the
least
votes, that should give a somewhat (not perfect) workflow.
Hope that makes sense,
Lodewijk
I stand on my point that it would be hard. :-) Not really in the tool supporting so many "juries", which would be simple, but in processing that. You could easily augment the data points by storing which images where viewed by a single user and if it was +1, -1, or skipped. We could also assume that all members are honest wikimedians and nobody is trying to game to contest. You end up with a pool of images ranked (eg. 1-10) by 1,000 different users. How do you get the top-10/100/500 images? The image some gave 10 points to, would barely have received 5 by others, and viceversa...
Wiki Loves Monuments mailing list WikiLovesMonuments@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikilovesmonuments http://www.wikilovesmonuments.org
Wiki Loves Monuments mailing list WikiLovesMonuments@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikilovesmonuments http://www.wikilovesmonuments.org