Commons:Wiki Loves Classics 2019/selection
Selection process for Wiki Loves Classics 2019
Preparation of categories
[edit]Revision of categories
[edit]--Epìdosis 19:00, 30 September 2019 (UTC)
- I think we can sign when we have completed one passage (as above). --Epìdosis 20:46, 1 October 2019 (UTC)
Spain
[edit]- Uncertain about some photos of d:Q250523, because the central part of the bridge is not Roman (only the arch at one end is Roman); in my opinion the photos depicting only the central part should not be included. --Epìdosis 21:13, 1 October 2019 (UTC)
- We can keep them, at least for now. --Epìdosis 21:26, 3 October 2019 (UTC)
- Are we interested in including dolmens (e.g. d:Q25475055/d:Q50689407)? --Epìdosis 21:17, 1 October 2019 (UTC) Or megalithic monuments of the Basque Country (e.g. d:Q66112650/d:Q66121625/d:Q66112464/d:Q66112443/d:Q66112567/d:Q66112444/d:Q66112455), complete list eu:Zerrenda:Arano-Goizuetako estazio megalitikoko monumentuak? --Epìdosis 21:28, 1 October 2019 (UTC)
- Images of d:Q2892716 should be well checked, there are Roman remains mixed with more recent rests. --Epìdosis 21:35, 1 October 2019 (UTC)
- If we are interested in the Phoenicians, there are 54 more images of d:Q3094251. --Epìdosis 19:13, 2 October 2019 (UTC)
- Done added. --Epìdosis 21:26, 3 October 2019 (UTC)
- User:Epìdosis we are including all "mixed" artifacts and buildings (maybe with a special mention "best mixture of old and new") and all ancient civilizations. The general idea is to be tolerant in order to experiment, it's actually complex to draw a line in may cases and it requires an expertise we sometimes don't even have yet. In case of doubt, let's keep. We can make a final decision at the end of the selection, eventually.--Alexmar983 (talk) 21:10, 3 October 2019 (UTC)
- I have included so far the Basque Cromlechs. If there is some images that survive the round because it's nice, we can think carefully later about what we want to include in the final rounds.--Alexmar983 (talk) 20:15, 17 November 2019 (UTC)
- We should carefully decide anout Castllo de Sagunto. Spain is completed besides a decision on these images.--Alexmar983 (talk) 15:39, 18 November 2019 (UTC)
- @Alexmar983: I've already reported it on the 1st of October ;-) I think I've selected only files containing ancient remains, so it's probably OK. --Epìdosis 08:19, 24 November 2019 (UTC)
Ukraine
[edit]hello there! i think we missed (somehow) the information about this contest... we can populate the category, but we need a bit more time. can you please tell me the deadline? --アンタナナ 21:38, 4 October 2019 (UTC)
- アンタナナ this is an experiment, so don't worry. We informed as many national competitions as possible and we actually missed you originally in the plan, so we inserted you few days ago. There is no rigid deadline. We were going to populate the category ourselves but you can do it as well, we will revise it. But take it easy, i repeat, it's a general test, the plan is next year to do things in advance. ok? All the jurors are informed we will need them for a much better event next year, for example. Would you like to suggest a juror for 2019 (and 2020). please let us know. Also, please tell interested uk-N user to enroll on meta: m:WikiClassics User Group.--Alexmar983 (talk) 22:25, 4 October 2019 (UTC)
- アンタナナ Done I've just finished to select all the images related to the contest, although I may have missed some, because of the high number of images and of my inexistent knowledge of Ukrainian language. Good night, --Epìdosis 23:37, 5 November 2019 (UTC)
Switzerland
[edit]We forgot Switzerland, I spoke to them in the afternoon, they entered WLM last minute. there is not a lot of things, I put for the moment the exterior of a museum and a bridge that was reconstructed on an old Roman bridge (since we discussed about a category for "mixture of old and new"), but it was worth to look there. I will put it in the table here soon.--Alexmar983 (talk) 22:31, 4 October 2019 (UTC)
- @Alexmar983: It's OK, I added Switzerland in the table and I've found also some other monuments. I agree about the exterior of the museum, while I think that the bridge, being totally reconstructed (I think there is no Roman material at all), can not be considered a "mixture of old and new" and should therefore be removed. Bye, --Epìdosis 22:00, 13 October 2019 (UTC)
Germany
[edit]d:Q34561254 (Zwölf Apostel (Langenbach)) could also be vorgeschichtlich as de.wikipedia states. Should we keep these images just in case?--Alexmar983 (talk) 17:14, 17 November 2019 (UTC)
- @Alexmar983: I would keep it. --Epìdosis 18:38, 18 November 2019 (UTC)
- ok, good. I can complete BY and therefore Germany.--Alexmar983 (talk) 18:55, 18 November 2019 (UTC)
All done
[edit]All Done, finally Category:Images from Wiki Loves Classics 2019 contains 3229 images. Thanks to Alexmar983, DarwIn and Geraki for the great job! Tomorrow we start with the first round. --Epìdosis 15:36, 24 November 2019 (UTC)
- Actually is this evening... but fine. I add here 90 more petrogliphs so the total is 3320.--Alexmar983 (talk) 15:38, 24 November 2019 (UTC)
Other categories
[edit]As I did with Category:Images from Wiki Science Competition 2017 we should create specific categories for valued, quality and featured images.
Waiting for the parallel sorting I think we can insert a gallery with those and monitor them (revising the categories, we will find them), it should give an idea of the level of possible finalists.--Alexmar983 (talk) 00:32, 10 November 2019 (UTC)
Evaluation
[edit]Step 1
[edit]- it s the usual "round 1" on Montage (yes/no phase)
- The 3,320 files (from 315 uploaders), all copied in the category, are sorted using Montage. They are imported in the tool with no filter. Eight jurors are involved in this preliminary phase with the thumbs up/thumbs down selection.[1]
- Jurors of step 1: Alexmar983, Epìdosis, DarwIn, Camelia.boban, geraki, Neuroforever, Ivana Cerato.
- The goal of this selection is to mostly remove images that are clearly unsuitable: poor resolution, people and objects obstructing the view, blurry composition and so on (and few files categorized by mistake in the competition). The core quality selection is scheduled at the next step.
- The reasons of this selection is as follows:
- we are doing a test for this first edition, and we want to understand what is actually there, this way we can estimate precisely the quota of informative files before focusing on their aesthetic quality. We can also estimate the quota of uploaders of only bad files, or if there is a mixture of different quality from the same person (which might mean that some uploaders can be encouraged to focus more on this aspect)
- the files are supposed to be polished, so it is important to see them again during different passages to perform this task. The more they remain, the more we can improve their metadata and categories and assess the aspect we want to improve for next edition.
- the proposed categories of our expected finalists (coexistence of modern and classic, museum collections, landscapes and views, writings and epigraphs) are not aligned to the standard winners of WLM, with only one category ("landscapes") which perfectly fits the type of files WLM usually selects. Since many of these images are not always at the top of a unified ranking, before discharging more of them we need a reliable overall view to select the most voted among those files. We need more jurors to see more of them before creating such reliable ranking in the next rounds.
Step 2 and 3
[edit]- it is the usual "round 2" on Montage (rating phase) but it's perfomed twice. This way a very smooth final ranking is provided.
- The "step 2" is in the end still part of the sorting phase. Normally, this could be done with a more rigid "round 1" but this was not the best strategy in our case. That's because, as pointed out in the previous section, our categories are not the standard categories of Wiki Loves Monuments and jurors have to gain expertise about them and also on the composition of the sample. Asking for a rigid selection by three random jurors would have produced probably too many false negative and positive.[2]
- The 1,697 files (from 263 uploaders), selected from step 1 (675 3/3 and 1022 2/3), are evaluated using Montage. Six juror vote per photo.
- Jurors of step 2: Alexmar983, Epìdosis, DarwIn, Camelia.boban, geraki, Neuroforever, Ivana Cerato, Richard Nevell.
- Results of step 2 are listed here (ordered by decreasing value of their average)
- The "step 3" is in charge of a smooth ranking. All jurors will vote now each photo of a much smaller set, carefully evaluating all of them. The quota of jurors from the academia increases. The scale of the previous step is now stretched taking in account more aspects, and jurors are encouraged to open every file for an accurate look.
- The 265 files (from 98 uploaders), selected from step 2 (all file above 5.83/10.0), are evaluated using Montage. Ten juror vote per photo.
- Jurors of step 3: Alexmar983, Epìdosis, DarwIn, Camelia.boban, geraki, Neuroforever, Ivana Cerato, Richard Nevell, Aleksandra Sulikowska.
- Results of step 3 are listed here.
- The order changes from the previous step because the number of jurors per image has increased and there is much more attention to the categories of the competition, so they are all more represented at the top of the rank.
- The most voted images are scrolled and a decision about the threshold of import was taken, aiming at the presence of at least two or three files per category in the final sample. Since this was not possible within a reasonable quantity (round 3 cannot be performed efficiently with too many files), an additional step is added (see below)
Step 4
[edit]- It completes the "round 2" with a selection of images with a more balanced presence of all categories. Specifically, duplicates of landscape images are drastically reduced, except of the case of images with high ranking and clearly different features (composition, moment of the day, presence of human beings, black and white).
- The most voted 94 images of the previous round (above the average of 6.75/10.0) are reduced to 50. here.
- Done by User:Alexmar983 in agreement with the jury.
Step 5
[edit]- It is the usual "round 3" on Montage (ranking phase)
- The 50 files (from 27 uploaders), selected from step 4 are voted by 13 jurors. Results are listed here
- Jurors of step 3: Alexmar983, Epìdosis, DarwIn, Camelia.boban, geraki, Neuroforever, Ivana Cerato, Richard Nevell, Aleksandra Sulikowska, Silvia Orlandi, Tyson Sukawa, Fabien Bièvre-Perrin
Step 6
[edit]- Validation by all the jurors and last refinements.
- A runner-up was also selected, since differences are limited in the average, Similar sites in the second position are skipped to maximize diversity.
- ↑ it's worth noticing that in Ancient Rome the meaning was in fact reversed than nowadays. :)
- ↑ This is still a problem on Wiki Loves Monuments sometimes even if jurors are quite expert, it could have been an issue in this case. The "double second round" is however not a novelty: it was used also for WLM2019 in Portugal and it shows it can produce very reliable ranking.