Talk:Wiki Loves Monuments brainstorm

Location

I'll probably come. Presumably where it says "Manchester" that is old information? 86.6.26.208 20:50, 10 January 2012 (UTC)

Thoughts on classes

Following on from conversations at the Jan 17 events, mentioned on the UK mailing list, there seemed to be considerable agreement that, apart from the general aim of recruiting people etc, the most useful actual photos might well be in specialized areas where we were weak, as opposed to general "old buildings" where we are mostly very strong. Such specialized requirements could be encouraged by special classes with their own prizes. Those mentioned, as I recall, were:

• War memorials
• Wrecks (there are "Scheduled wrecks")
• Church furnishings
• Side of a building murals
• Historic defence works (what's the site for that?)

Please add any other thoughts above or below. Johnbod 14:04, 19 January 2012 (UTC)

I agree wholeheartedly with Johnbod that we should try to concentrate on specialized areas where there is the opportunity to do something useful that doesn't just duplicate what is already available elsewhere. My particular interests would be in:

• Ogham stones -- I uploaded images of 4 stones from Cornwall and Devon a while back, but there are 30+ stones in Wales, nearly 40 in Scotland, and half a dozen in the Isle of Man, and most of these are in situ.
• Anglo-Saxon crosses, especially Runic crosses

BabelStone 14:37, 19 January 2012 (UTC)

Just to have a clear image: with focus, do you mean give people an extra incentive to upload those images, or to forbid them to upload other monumental sites? The first is definitely a good idea as long as it is well communicated (picking one single topic as focus might help), but it seems that negative incentives (forbidding) usually don't work as well. For example, we didn't forbid people to upload images of the Eiffel tower, although we didn't need them - because those people might then after that also upload more useful images because they learn it is fun and welcomed. Effeietsanders 14:54, 19 January 2012 (UTC)

I think we were thinking in terms of special classes with their own prizes for some of these (we don't have to do them all), but also a general class for, let's say, all listed buildings. We were also thinking of prizes for categorizing the uncategorized stuff we already have - an unknown 6-figure number of images. Johnbod 22:10, 19 January 2012 (UTC)

I just talked with Maarten, and he mentioned that he has hopefully a bot standing ready to at least take a significant part of these six figure numbers and categorize those - apparently a lot of that can be done automated (he did it for a first load as well). So before we decide anything on that lets wait a bit to see how that goes. But good to know you were indeed thinking about 'special focus' and not exclusion! Effeietsanders 22:24, 19 January 2012 (UTC)

Actually categorizing them might be overambitious - Multichill put a lot of work into this but, perhaps partly because of errors in the original metadata, the error rate was unacceptably high and he gave up. Merely correctly identifying which ones actually are uncategorized would be huge progress. Part of the problem is a template identifying them as uncategorized, which persists after they are categorized by Cat-a-lot (?and Hot-cat). Werespielcheckers is the expert here, and has been talking with Fae about a bot to help here. At the moment if you search in Commons on a location typically 25% of the results don't have a good location category (never mind antything else) but 75% do, adding greatly to the hassle of trying to add categories. Johnbod 16:41, 20 January 2012 (UTC)

The good news is that cat a lot has been greatly improved in two different ways during the last few weeks, and this should enable existing and new categorisers to speed up the process. The category box now has predictive texting similar to Hotcat, and it also now removes the geograph uncategorised template when you use it to copy images from a geograph category such as a geocode one to another category. Hotcat already does that, but Catalot is point and click whilst with Hotcat you have to open each image. So some geocodes are going to be readily categorised using catalot and many thousands more images are likely to be categorised by September.

The bad news is that the bots are giving us some intermediate categorisation and when they've added categories you have a partially categorised geograph template that only comes out if you edit the image. Fae and I have been discussing getting a Bot to remove the partially categorised template, but I suspect it would need to be a bot that identified images that have also been categorised by Hotcat or Catalot.

Some of it could be improved by refining the bots, but ultimately it often needs a human to avoid stuffing huge numbers of images into high and mid level categories. I wasn't surprised at the size of some of our categories in touristy areas, but 3,500 images in the category Strabane did look to me that it needed more subcategories.

It would only take a small number of extra categorisors to do thousands more images but we have 900,000 uncategorised Geograph images and getting more than a few percent done in 7 months is a tall order. Frankly whether we get 3% or 13% of it done before WLM starts in the Autumn we will still have a humongous categorisation backlog when the normal time comes for Wiki loves monuments. Hence my previous suggestion that we encourage participants to search commons first and look for gaps to fill in. WereSpielChequers 18:34, 11 February 2012 (UTC)

May 2012 - after AGM

I've uploaded my slides with an added link to this section for comments, white knights & volunteers interested in any aspect of this. Johnbod (talk) 21:46, 15 May 2012 (UTC)