User:MargaretRDonald/sandbox/Thoughts on judging wikilovesearth

Easy enough in many respects to judge images, but ultimately the purpose of the competition was and is, to generate photos of high quality which can be used to illustrate Wikipedia articles. Hence, every photo had to be examined with the aim of possibly inserting it into an article, if not several. All photos had already been sorted into the "best" from their country. But many were only captioned and described in their language of origin, with the most informative Commons category being "National park of Malaysia" or "Protected area of Rwanda" or some such. If they were Taiwanese, they came with a link to a description of the protected area, so that one had a name for the province or perhaps a bit more. And then many were titled in Tamil, Belarus, Armenian, Chinese and various other tongues using different orthographies, with no English captions or descriptions. I resorted to Google translate.

Hence there was a lot of work to be done in Commons writing English descriptions, finding and sometimes creating categories imitating others thus Category:Unidentified spiders in Bolivia And lots of work in Wikidata with so many national parks, protected areas being labelled in their own language (Tamil, Chinese, Armenian) but having no English label.

Many of the images from Brazil did not have appropriate national or state park articles, and when the image was that of an animal it generally was not identified either by common name or genus or binomial name. I think it could be useful to try to find some experts to see if they can identify things in Commons. Certainly many of the Brazilian Commons items. See, for example, Uatumã Biological Reserve which was distressingly less than useful.

Articles are missing in every Wiki for RAMSAR sites and other protected areas.

Identifications
There was a common problem of images lacking sufficient identification. Where these were essentially landscape images, the name of the specific park or protected area was often difficult to work out. However, many countries, did pre-process images to identify the specific protected area. More problematic was the non-identification of plant and animal species when they were the subject of the photograph.

Pre-processing images
Countries often pre-processed images to include categories such as Category:Protected areas of Thailand, and to provide links to description of them as for example. Countries which provided links included Latvia, Spain, Thailand, Taiwan, Ukraine, Uganda, Portugal, Canada .... Sometimes all that was available was e.g., Category:Protected areas of Brazil. This pre-processing is a vital component in any Wiki Loves Earth campaign, since a majority of the images came from protected areas.

Parts of the description linking to external sites

 * This is a photo of a natural heritage site in Thailand | 1001
 * This is a photograph of a Special Area of Conservation in Spain with the ID: ES6140004. Natura2000 entry, EEA entry

Statistics
I was asked to judge just 337 images. However, an analysis (using GLAMorgan) shows that if we were to consider the month of September, 2020 and the images uploaded to "Wiki Loves Earth 2020", then


 * 4032 files were viewed out of 4738 inserted into articles
 * 6080 pages on 207 wikis use these images
 * 5,112,138 file views in 2020-09
 * (608 pages were unable to be uploaded to the GLAMorgan analysis)

Recommendations

 * 1) Present each judge with a different randomised set of images. This may already be happening, but even better judges than me, will not be consistent over time. This issue is simple to test statistically, and such testing should be done if it hasn't already. (My judging was not consistent over time which may not matter all that much. I suspect I may have been consistent in giving the highest rankings - but rankings less than 8/10 were not consistent over time).
 * 2) Pre-processing is critical. That is we must try to ensure that:
 * 3) protected areas are identified (either by external identifiers, and /or more adequate categorisation). The South American protected areas have many photographs from tourism, but in general these are very poorly identified. (There were two issues here: articles for such parks not yet written, together the non-identification of parks and animals.)
 * 4) Find a process to usefully categorise animals & plants