[01:17:29] What's the rough ETA on Abstract Wikipedia? [03:24:59] Wikifunctions this year, Abstract Wikipedia hopefully the next [03:33:48] 。 [14:01:35] when posible, someone please explain better the difference between Abstract Wikipedia and Wikifunctions in https://meta.wikimedia.org/wiki/Abstract_Wikipedia/Overview [14:09:03] the title of that page is about Abstract Wikipedia, and the text is about Wikifunctions, so I thought they are the same thing, but I see in the vrandecic's comment above they are not [14:13:09] the difference is stated better at https://meta.wikimedia.org/wiki/Abstract_Wikipedia#Project but I guess some efforts to better define the two projects there can be made [14:59:02] I am curious about how the "abstract articles" will be translated to each language, will we need to create Lua modules to do that or there will be some extension to translate or some other way? How the Wikifunctions functions will bind to a Wikipedia article? [15:08:40] danilo: Most of the content on Meta will move to Wikifunctions.org once that exists, at which point there will be dedicated sets of pages for what Wikifunctions is about and what Abstract Wikipedia is for, and not a single page set covering both. But until then it's a bit confusing, yes, sorry. :-( [15:09:59] * James_F danilo: To answer about abstract articles, the location of where abstract definitions will be written and where they will appear is yet to be decided, but the idea is that the Wikifunctions-powered community-controlled systems will do the translations into languages; you won't be writing local Lua code or whatever to translate them outside of the system. [15:13:06] danilo: As an early step, we're considering exploring how abstract content could be used to provide the short descriptions on Wikidata with much less work for the community. Some thoughts are in the update from the end of July 2021: https://meta.wikimedia.org/wiki/Abstract_Wikipedia/Updates/2021-07-29 [15:15:19] As another early step, there is already an early prototype for generating certain types of short descriptions in German: gitlab.com/mahir256/ninai/-/blob/main/demonstrations.md (section "Descriptions of places and people", h/t Nikki) [15:16:22] James_F: thank you for the explanation, I will read that [15:43:38] another question, is there some study/discussion about how will be the impact in local Wikipedia communities of the add of thousents of abstract articles in the Wikipedias? [15:44:24] I don't think so, but it's up to a given language community whether they want to use abstract articles in addition to handtyped ones [15:46:07] Yeah. Given that the technology for creating abstract articles won't exist for at least a couple of years at this point, it's probably premature to worry too much about how each community might want to think about adopting it. [15:47:17] If you want to hasten this worry, join me in developing Ninai (the tool whose demonstration page I linked earlier)! :-) [15:48:48] the local comminities will can choose the abstract articles they want one by one? [15:48:58] Precisely [15:49:50] hmm, so I think probably there will not be too much problems [15:50:14] * James_F We don't even know where the articles will "appear", as we've not had that discussion yet. [16:10:24] one of the first impacts I can think is that some Wikipedias will change the local version of stubs of species, cities and astronomical objects to the abstract version [16:11:26] those articles are often created by bots, watched by no editors and are hard to keep up-to-date [16:13:13] giving it to another commuity take care can reduce some of the manutence load of the local Wikipedia community [18:04:42] Yeah, that's a good point. [19:44:00] here's the same code re-run in a Python terminal just now : https://tools-static.wmflabs.org/bridgebot/d4ff6027/screenshot_2022_02_02_13_41_22_994651898.png [21:31:36] are there also plans to create codes to make the translation from natural language to abstract? [21:32:30] that's not an explicit goal of the Abstract Wikipedia project [21:32:44] and besides, I don't think it is feasible without entering the realm of neural nets (something @vrandecic wishes to avoid due to issues with explainability) [21:36:30] wihout that 'inverse translation' we will need to write all abstract articles from scatch, with some code to make at least a 'pre inverse translation' it will be easier to create new articles [21:38:30] maybe that can be done with a tool in toolforge [21:38:49] Yes, but thanks to Wikidata, we can write groups of articles at the same time. As with your example, we can start with generic abstract articles about species, cities and astronomical objects. And then we can refine them over time. (re @wmtelegram_bot: wihout that 'inverse translation' we will need to write all abstract articles from scatch, with some code to make at le...) [21:39:41] I can recommend this interview I did with Denny where we specifically touch upon that: https://wikipediapodden.se/episode-154-abstract-wikipedia-and-wikifunctions/ [21:40:16] (not sure if my screenshot made it over the IRC bridge, but) within it you can see that there are three Wikidata Qids in each line. substituting those with (an item, its P31, its P17) or (an item, its P106, its P27) is a great start [21:56:05] I understand, there are ways beter than a inverse translation to create the first articles [21:59:00] but I also think some kind of inverse translation tool could help those who want to create or improve articles but think the 'abstract language' to complicated