(Reported on GitHub by @Lokal_Profil)
Currently it looks like whover uploaded the latest version gets credited as the Photographer. E.g. for Valö2, Vrångöskärgård.jpg We see that Embedded_Data_Bot is credited instead of Yvalima.
As this also affects participant and new user counts it can potentially skew the statistics.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | Lokal_Profil | T172772 Wlm/heritage/erfgoedbot sprint at Wikimania Hackathon 2017 | |||
| Resolved | JeanFred | T173003 Ensure wikiloves tool detects first uploader, not latest |
I deployed and rescheduled a database update. Here are the differences:
| Year | Countries | Uploads | Images used in the wikis | Uploaders | Uploaders registered after competition start |
| 2013 | 1 | 9691 | 1764 (18%) | 350 | 276 (78%) |
| 2014 | 13 | 60439 | 8663 (14%) | ||
| 2015 | 25 | 104319 | 11743 (11%) | 8862 | 7732 (87%) |
| 2016 | 24 | 110686 | 9376 (8%) | 13297 | 11809 (88%) |
| 2017 | 35 | 132070 | 8070 (6%) | ||
| Year | Countries | Uploads | Images used in the wikis | Uploaders | Uploaders registered after competition start |
| 2010 | 1 | 11983 | 7617 (63%) | 202 | 141 (69%) |
| 2011 | 18 | 156734 | 67521 (43%) | 5541 | 4119 (74%) |
| 2012 | 34 | 352232 | 107617 (30%) | 15279 | 12296 (80%) |
| 2013 | 53 | 363870 | 93382 (25%) | 12461 | 9577 (76%) |
| 2014 | 41 | 305687 | 52705 (17%) | ||
| 2015 | 32 | 199729 | 37492 (18%) | ||
| 2016 | 42 | 275256 | 37265 (13%) | ||
207 uploaders --> 202 uploader (same newbie count)
It’s a bit all over the place... It’s not exactly clear to me when the last database update had actually run (it’s hinted it was only Aug 1st) so just the time difference may explain the differences.
(There might also have been plain bug fixing.)
cc @Effeietsanders @LilyOfTheWest @ezachte @Lokal_Profil for information
I’m not actually sure it worked :-( There are still eg SteinsplitterBot in http://tools.wmflabs.org/wikiloves/earth/2016/Pakistan :-(
If we adapt the Polish competition (or WikiDaHeim) to be included in WLM as they should anyway, we could check those as test sets :)
Hmmm. The harvesting process had quite a few hiccups recently − had to poke it quite a bit to make it work. Maybe the update just had not run...