Jump to content

User:MargaretRDonald/sandbox/Thoughts on WLE 2021 In Australia

From Wikipedia, the free encyclopedia

This consists of my hopes, technical discussions, recommendations and the reasons for them.

Hopes

[edit]

I had hoped for:

  1. tens of thousands of entries, from
  2. many hundreds of new users (who had made themselves accessible)
  3. species images for animals and plants as yet unrepresented in the commons (4 or so images fell into this category)
  4. images from indigenous protected areas
  5. species and habitat images from ecological communities
  6. images from professional botanists which would fill some of the gaps
  7. images from far-flung places as yet unrepresented in the commons.

To these ends, I contacted at least 3 IPAs, and many professional botanists. It is unclear whether any of these responded. Hope 7 was the only substantial hope to be realised. It was enormously satisfying to be creating new wikidata and new wikicommons entries for so many images.

We received 1510 images from 281 users, 261 (93%) of whom had never contributed before.

comparisons

[edit]
Country Landscape Detail Total
Ghana 217
Azerbaijan 229
Croatia 329
Indonesia 678
Benin 927
Greece 930
Australia 304 1510
Turkey 2446
Armenia 6124
Ukraine 11286
Germany 11747 1950 13702
State representation
State   # contributors # images Proportion
SA 34 56 4%
TAS 21 50 4%
NT 31 138 10%
VIC 40 212 15%
QLD 40 206 15%
NSW (inc ACT) 87 259 18%
WA 50 498 35%
Viewing and usage statistics for 2021-07 for the Catgories:Images from Wiki Loves Earth (year) in Australia
Year files viewed used %files used pages # wikis file views fileviews/upload file views /page
2016 4067 334 /385 9.47% 1402 93 1020871 251 728
2017 1503 109 /122 8.12% 247 52 245177 163 993
2018 987 75 /85 8.61% 122 33 68886 70 565
2019 698 49 /63 8.88% 33 29 88993 127 2697
2021 1532 112 /138 9.01% 278 32 25841 17 93
Year Country files viewed used %files used pages # wikis file views fileviews/upload fileviews /page
2021 Armenia 6125 47 48 0.78% 39 15 67079 11 1720
Spain 1323 19 20 1.51% 69 31 52239 39 757
Portugal 482 9 9 1.87% 7 2 24600 51 3514
Indonesia 680 13 16 2.35% 21 10 4600 7 219
Italy 1748 41 42 2.40% 52 12 38148 22 734
Benin 949 24 24 2.53% 14 3 12606 13 900
Azerbaijan 231 5 6 2.60% 10 7 519 2 52
Nepal 585 20 20 3.42% 16 5 52513 90 3282
Ireland 362 9 14 3.87% 16 4 1025 3 64
Russia 6216 336 338 5.44% 191 20 1455949 234 7623
Rwanda 311 18 18 5.79% 24 6 23925 77 997
Germany 13734 746 922 6.71% 703 17 207559 15 295
Ukraine 11285 784 790 7.00% 401 4 68945 6 172
Sweden 1825 281 329 18.03% 405 3 14585 8 36
Austria 457 117 131 28.67% 148 46 82817 181 560
Philippines 1012 296 303 29.94% 333 32 122324 121 367

Failures

[edit]

To persuade contributors

  1. to read (and follow) the rules
  2. that one of the main purposes in uploading an image was to make a contribution to an encyclopaedia
  3. to give as full a description of their image as possible
  4. to pick their best from a series of images.

This means that next year I want the (Australian) rules & messaging to be carefully reviewed.

Messaging within wikis

[edit]
  1. Banner message & banner. (Next year I would like an Australian banner)
  2. Message on the landing page
  3. & within the subpages

These were changed at various times throughout the competition, in an attempt to get images which satisfied the criteria. This was hardly ideal. I plan next year that these messages will be reviewed and agreed to and therefore hopefully not need changing. In some respects using European models for these led me astray: the Swedish message may well attract the kind of desired entries for Sweden. For the brief period I used it, I believe it was a mistake for our Australian competitors.

Recommendations

[edit]

Major recommendations

[edit]
  1. The organiser of WLE should be a member of the wikimedia Australia committee for the duration of the competition and possibly up to two months prior to its commencement. The 2021 competition suffered from this omission: there was no regular means of oversight, or communication to allow the necessary discussion.
  2. The organiser should serve for two years (at least) so that fewer things are an unknown surprise.
  3. The competition be organised by a committee of at least two: the organiser, and the organiser elect (whose role would be to advise and check on all technical aspects of the competition, and be a sounding board, and an advisor).

Minor recommendations

[edit]
  1. That judges not be paid. An enormous amount of work was required behind the scenes, once a competitor had uploaded a photo. The work was done very unequally and mostly not by judges. Remunerating judges seems to me unfair to those who did all the hard yacker and yet the nature of wikimedia is that we all contribute unequally. I feel that this precedent should not be repeated as it upset me deeply that those who had shouldered the burden of the work were those who were not remunerated. (I also find it contrary and contradictory in terms of our movement.)
  2. that we issue media releases in all states and territories in an attempt to get mainstream media coverage (as well as social media) for the competition. Alex, Caddie. Can we measure our outreach on social media? I believe it was fairly minimal. I know of only three of us who tweeted just a few times, and although some of those were retweeted but I suspect to the same small community. I did a radio interview on WA ABC radio (very brief, and frustratingly not to the point).
  3. that we monitor (ideally set up) the competition on a state-wide basis. This could engender competition between states and hopefully more images. It would be great to copy the German federal set up, but this is in my view too ambitious for next year.
  4. That we have prizes for
    1. peoples' choice. We can run the prejury tool as the Germans do, until the time for submission to the international jury. We can set the cull for the judges on any predefined date (this year 30 June 2021). We can and should perhaps run a banner inviting Australians to vote (using the prejury tool). If we were to attract numbers of voters in their hundreds by use of mainstream and social media, it would make sense to give a people's choice prize. This would encourage participation and interest (hopefully).
    2. organiser's choice This would be a very kind recompense to the organiser, who laboured so hard. The photograph need not necessarily satisfy the international criteria.

Organiser's choice for 2021

[edit]

These images contributed to multiple wikis and answered in my view a major purpose of the competition. None were able to win (two because they were too small and and the remaining one was not of sufficiently high quality).

  1. Tacca maculata
    Tacca maculata in Mirima National Park
    There had been no article for this, and Shib68 gave us two images, taken in the Kimberley in Mirima National Park, for which he/she contributed several other useful photos to the category.
  2. Enteles vigorsii – a beetle which had no Wikidata entry. The photograph is poor but it led to a Wikidata entry, a category and an article. Invertebrates are particularly poorly represented in Wikidata and Wikipedia. We need to make a push for invertebrates and, in particular, to at least get most of the Australian Faunal directory IDs up on Wikidata.
    Enteles vigorsii in Lamington National Park
  3. Nototodarus gouldi in Bremer Marine Park
    Gould's squid (Nototodarus gouldi) This is the only photograph in Commons of this species but there are two superb illustrations by Elizabeth Gould. The quality of the photo is poor because it has been extracted from a video. This was part of a wonderful series by Jessica Meeuwig taken from Baited remote underwater videos in surveys of the marine parks off the coast of Western Australia.

Things I would do differently next year

[edit]
  1. Use the German prejury tool on a statewide basis to make less work for individual jurors. This tool would continue to be used because unlike the Ukrainian jury tool it adjusts to incoming images.
  2. Set the initial cull to 200
  3. run the first round with the final judges as a thumbs up/thumbs down rating across 200+ images(dependent on cut point). This would use the Uraininan jury tool.
  4. Use the 1-10 rating using the Ukrainian jury tool.

Technical

[edit]

"Prior" to the competition

[edit]

The following pages on commons all formed part the technical setting up and running of the competition. In some cases, they suffered important changes during the competition (Such changes are not desirable but were necessary in the view of the organiser and perhaps might have been obviated with more communication with the committee)

  1. Participating countries in WLE2021 (This commons page was set up by the international organisers and completing the row for Australia permitted our participation in WLE for 2021)
  2. Wiki Loves Earth 2021 in Australia (This was the commons "landing" page for competitors after clicking on the banner. It led them through the process. This page, too, changed multiple times during the competition.)
  3. Category:Wiki Loves Earth in Australia (commons: Page for the uploaded photos)
  4. Commons:Wiki Loves Earth 2021/CentralNotice The central notice gave the words for the banner used to attract wikipedia users' attention to the competition. The wording was changed at least twice during the competition.
  5. Competition images This page was a requirement of the international page.
  6. Australian Prejury tool A link to the German/Australian prejury tool. An innovation in 2021. We terminated this on June 30, but next year would prefer to terminate it close to the final judging to allow a people's choice.
  7. Australian Prejury tool: created selection for jury (admin choice from the prejury tool -accessible only after login) A link to the prejury tool allowing the organiser to see results from the rating process.

During the competition

[edit]

Each image (almost without exception) required that we:

  1. assess its eligibility: by checking
    1. its size; and whether
    2. it was taken in a protected area
    3. it satisfied licence requirements. This was checked only for the final images, via Google images reverse search and we found a number of images in the final group for judging which needed OTRS licencing. The reverse searches were undertaken after the competition, but probably should start to be undertaken for any photo that during the competition (given the number of problematic images.)
  2. attach "depicts" statements to allow international judges (German, Bengali, Thai, Chinese, Arabic speaking, Nepali) to recognise that the image was taken in a protected area. This work frequently involved the creation (and filling out of) new wikidata items
  3. attach meaningful categories to the images. Again, this frequently involved the creation of new categories (and new wikidata items linked to the commons)

Images that failed the eligibility criteria were put into the Category:Images from Wiki Loves Earth 2021 in Australia outside the competition. This was an enormous amount of work, which lessened the number of prejury images to look at. However, by the end of the competition, I took the view (and advised others) to simply rate images outside the eligibility criteria with a rating of 1. This would mean that they would never have to rate them again and that they would probably then be excluded in the first cull based on their average rating. In addition, I knew by then, that the final jury tool allowed me to set a blanket exclusion by size.

The massive task of adding "depicts" statements, wikidata items, categorising and adding categories for the majority of the entries was strictly only necessary for the images to be submitted to the international competition. However, in fairness to our new contributors and to the possibility of their images being used across wikipedias, it was a necessary task.

The majority of the competitors were not capable of putting up either useful "depicts" statements or useful categories.

Other tools

[edit]

Some useful pages

  1. Organising WLE Monuments 2020
    1. Poidabro's invitation to participate in survey for WLE 2020 It would be good to survey our competitors.
  2. Reverse image search (a necessary check for the images uploaded to the international competition)
    1. tineye This failed to work effectively. It failed on URLs and on jpegs did not find occurrences of images used multiple times on the internet.
    2. Google images reverse search This failed to work for URLs, but worked for uploaded jpeg files.
  3. Petscan query with uploader, date uploaded, imagename and other information about the Category:Images from wiki loves earth in Australia. A CSV file will be saved when, under "Output", and under "Format", "CSV" is selected and the "Do it" button is pressed. This was used for summarising uploads.
  4. Dashboard for WLE contributors
  5. Jury Tool

Post competition

[edit]
  1. Judging: This year we used categories plus spreadsheets. However the Ukrainian jury tool (after nearly four months) was used for the final round. Next year I would hope to use of this tool much earlier in the competition. However, it is still the case that the Australian organiser is not able to do the initial setup. Hopefully, next year it will be available to us prior to the closure of entries. In which case it could be used for the cull, as it has the possibility of ratings (0-1) The (0-1) rating could be used for the cull and has some advantages over the German prejury tool)
  2. Renaming: Many images do not have useful titles (or descriptions). When we have informed the international competition of our 15 photos, all items whose names need changing will be renamed. For example, DSC cfg.jpg is not found on a simple search. It needs to be renamed, as do most of this contributor's images. These should not be changed during the competition, because the prejury tool records recognises a renamed image as two entities.
  3. Incorporating: A push to incorporate images into articles would be fantastic. This is more easiily done as images come in. I was very pleased to see the #WPWP campaign follow close on the heels of WLE2021.

Jury tools

[edit]

German prejury tool

[edit]
Number of ratings / image

This graphs shows that the images presented by the prejury tool were randomised. (Had they images been presented as first uploaded then we would expect some images to have almost 47 votes)

  • Uploaders were excluded from casting votes for their uploaded images
  • Culled to ~100 images using German prejury tool (47 raters)
  • 47 raters who cast 16152 votes (votes/rater=437)
  • For the cull the cutoff of 3.1818 was chosen giving 101 final images
  • minimum number of ratings given for final images left from the cull was 11



Ukrainian jury tool

[edit]
Usage of the Ukrainian jury tool for a round
  • Culled to 29 images on +/- scale (5 judges)
  • Used 1 to 5 ratings on 29 images (5 judges gave 3 groupings)