Nuno,
Thanks for the code. It looked promising, but unfortunately I'm not a
programmer and the final product never arrived, and I think we'll have
to "go manual" at this point.
To convert your system to our minimum requirements, we'd have to take
your "single person" system and
A. Randomly select, 50 or so photos, and have one person rate these, say 1-5.
B. Store these results
C. give a new screener another 50 or so photos, and store these results, and
D. do this so 100,000 photos get rated at least 3 times each, then finally
E. Average all the result for each pic.
Obviously some bells and whistles would have been nice, but that would
have worked. Unfortunately it just never got that far.
If anybody can deliver that in working order before Sept. 10, please
let me know!
Pete
User:Smallbones
------------------------------
Message: 3
Date: Thu, 30 Aug 2012 20:24:13 +0100
From: Nuno Tavares <nuno.tavares(a)wikimedia.pt>
To: wikilovesmonuments(a)lists.wikimedia.org
Subject: Re: [Wiki Loves Monuments] Automated voting/screening system
Message-ID: <503FBDDD.2000107(a)wikimedia.pt>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Hi Peter,
Didn't our code help? Could you be more specific why the solution
doesn't apply to you?
-NT
Em 29-08-2012 16:37, Peter Ekman escreveu:
It looks like WLM-US is having bad luck coming up with
an automated
voting/screening system to sort out the large number of photos we
expect to have uploaded. Does anybody have any suggestions or
technical solutions? Just to give the basics, I think WLM-US has to
be prepared to screen 100,000 photos down to the 500 that the jury can
deal with directly. That will likely take input from over 50 editors,
e.g. the general logged in US population of editors, and take
processing from early on in the contest while the contest is still
running, e.g. before Sept. 10. Any help appreciated,
Pete
User:Smallbones