Talk:Abstract Wikipedia/Updates/2021-07-29

You can discuss abstract descriptions here. --DVrandecic (WMF) (talk) 22:09, 29 July 2021 (UTC)Reply

It's a pleasure to have own bot make it to this newsletter. However, when you clicked "Random item", you most likely saw my bot importing labels, not descriptions. In particular, it is supposed to scan changes in the week before the last one, check for sitelink additions (eg., new Wikipedia links) and import new titles as a label in the corresponding language. If I should mention a bot, it would be Frettie's Frettiebot which created (Czech) descriptions for places, people and many other things quite some time ago. Perhaps he has some ideas to share.

Nevertheless, I did actually do some efforts on importing descriptions, but they are somewhat orthogonal. Some Wikipedia pages are disambiguations -- they navigate readers when the page name is ambiguous and link to other pages with the same name. They are usually structured like:

link - some human-generated description

I had an idea to scan these pages and import them by bot, but sometimes the descriptions are not optimal for Wikidata and it's better done supervised. So instead of direct importing I put them to a database and had people add/reject them using Magnus' (who by coincidence is also mentioned in the issue) Wikidata game.

My very recent effort is the exact opposite way, ie. mirroring Wikidata descriptions to Wikipedia disambiguations using a template and a Lua module. This could be a use-case for abstract descriptions.

Fellow editor Draceane even made some efforts in actual generating descriptions from data (code), which is quite close to truly abstract descriptions. --Matěj Suchánek (talk) 11:08, 30 July 2021 (UTC)Reply

I added a lot of descriptions in Wikidata and for preparation I use a Spreadsheet where I define how the description will look like. To make it understandable for people it is important that the notation is not so complicated. I am interested in abstract descriptions but I also think that with the current situation that many descriptions are added by bots some people have a task and it was something that I liked. I have spent a lot of time in the last 3 years with adding descriptions with QuickStatements. From my point of view there is a need that people can understand how a specific description was generated before it was added to Wikidata. If Wikifunctions can help with that it is something that I support. I think a possible way how propably many people can write Functions or define Descriptions is with Spreadsheets. There are in some languages specific differences in the descriptons as far as I have seen. For example sometimes some parts of the description are in brackets. For such things it is important that that it is possible to define the abstract description per language as I can define in my Spreadsheet it per language. The mostly used language style is different between the languages. So I am not sure if it is good to follow the same structure for all languages. I think it is important to accept it and see it as a challenge. With my work on converting Spreadsheet functions into source code in other programming languages I think it can be possible to use it for abstract descriptions if I understand the notation or someone else can help me. I have a version of the program that runs on the R-Studio Installation in Wikimedia PAWS but it is not complete yet. I have a more current version on my computer and will update the one in PAWS in two weeks during the Wikimania Hackathon.--Hogü-456 (talk) 20:26, 30 July 2021 (UTC)Reply

It's really great to see this new milestone set as it is going to materialize the idea of Abstract Wikipedia or rather Abstract Wikidata in this particular case. For years now I've been working on items of Polish scientists (and turning author name strings into authors via Author Disambiguator tool. I would like to draw your attention to descriptions of scientific articles. They form over 30% of all Wikidata items. Adding descriptions to them hardly requires human intervention, they all can be described as scholarly article/scientific article. Probably more info should be automatically added to further disambiguate the description like scholarly article published in 2020 or scholarly article published on 27th July 2020. Would it be an initial target for this milestone ? Kpjas (talk) 08:36, 31 July 2021 (UTC)Reply

The scientific articles descriptions are a good example and there are a lot of items about scientific articles in Wikidata. I added descriptions that follow the structure: in Month Year published scientific article. That have I done in German and so it is an translation what I wrote as the example. I for my self prefer in German that way of descripton with the mentionioning of the Date before the description what it is for scientific articles. In other languages I have seen descriptions with words in brackets about scientific articles. So it seems to me that the styles of descriptions are different and this is something what I think is good. As long as it is a correct short sentence I dont have a problem if all descriptions follow the same structure across languages. As far as I know there is for not logged in users the content displayed as English for search. So these description is the most important.--Hogü-456 (talk) 13:30, 31 July 2021 (UTC)Reply

I like this idea of Abstract descriptions for Wikidata. This will indeed be the first step towards testing Abstract Wikipedia and with a special language code (like qqz, as mentioned on the content page), contributors can experiment with short descriptions using ZIDs/renderers without breaking/removing existing descriptions. John Samuel 16:17, 4 August 2021 (UTC)Reply

@DVrandecic (WMF): This proposal is too hidden and masked under the page name "Abstract Wikipedia/Updates/2021-07-29". This page should be named "Abstract descriptions" and linked directly from the {{Abstract Wikipedia navbox}}. While the whole Abstract Wikipedia is more of a phantasmagoric utopia (or a completely new project completely different from the current Wikipedia), "Abstract descriptions" can start as a simple tool that could be up and running in a matter of weeks. Such a feature is much needed for Wikidata to reduce unnecessary multiplication and unjustified differences between descriptions in different languages, as well as very common errors, deficiencies and nonsenses that are now difficult to check, detect and treat.

"Abstract descriptions" and "automatic description" are two concepts that should be combined. The default abstract descriptions can be generated automatically by classes, statements and defined rules, but manually editable. Presentation of the abstract description as searchable (but non-editable) text in all individual languages is also an application of the principle of automatic description. --ŠJů (talk) 22:57, 26 June 2023 (UTC)Reply

Hi ŠJů. Thanks for reading and for your good suggestion. I've added a link to this page within the /Tasks page for now (Abstract_Wikipedia/Tasks#Task_P2.11:_Abstract_Wikidata_Descriptions).

We need to overhaul the navbox soon as it is getting overwhelmingly large, but will keep that suggestion (along with adding other key newsletter entries) in mind when we do so.

I imagine we also need to update some of the ideas here, now that 'mul' is getting closer to a testable state (phab:T285156 - they're currently working on milestone 0.4, per the latest Wikidata newsletter).

Thanks again! Quiddity (WMF) (talk) 23:46, 19 July 2023 (UTC)Reply