[09:36:06] I was thinking today of the contribution of Abstract Wikipedia to Wikidata. [09:36:33] It will be excellent to use Z-items to generate descriptions of Wikidata items. [10:17:04] For what? [10:26:10] e.g. {1} in {2} where {1} is the item class and {2} is the country. (re @amire80: For what?) [10:26:43] Mass generate descriptions: Monument in India, Theatre in Egypt... [10:27:36] we already have bots mass-generating descriptions, but a lot of people (me included) would really like to see them generated automatically instead of needing an army of bots updating items constantly [10:28:08] This is the point (re @Nikki: we already have bots mass-generating descriptions, but a lot of people (me included) would really like to see them generated automatically instead of needing an army of bots updating items constantly) [10:28:23] (it's kinda ironic how one of wikidata's first aims was to remove the need for an army of bots updating sitelinks, only to need an army of bots itself to update descriptions...) [10:28:51] :) (re @Nikki: (it's kinda ironic how one of wikidata's first aims was to remove the need for an army of bots updating sitelinks, only to need an army of bots itself to update descriptions...)) [10:29:41] I'd kill for bots that don't update things incorrectly in certain languages 😂 [10:30:09] Or IP addresses that don't 😁 [10:45:18] I mean, for what purpose. [10:45:19] What's the problem if an item doesn't have a description? [10:45:21] If a description can be generated by a bot, maybe it's not necessary in the first place? (re @Csisc1994: Mass generate descriptions: Monument in India, Theatre in Egypt...) [10:46:01] we use descriptions when editing to help us identify the right item [10:46:19] When editing what? [10:46:24] wikidata [10:46:54] I can understand that a description is needed when two or more items have the same label, but if the label is unique, is a description really necessary? [10:47:45] yes, because the right item might not have been created yet [10:50:33] Mmm... okay. I just hope that if AW is used for this, it's more automated and on the fly, and that it's not a bot that creates something using AW functions and then leaves it to rot. We essentially have that already :) (re @Nikki: yes, because the right item might not have been created yet) [10:50:49] yeah, that was what I meant :) [10:53:17] i'd love to see descriptions generated automatically when an entity is loaded and then cached, then they could be included in the json too so people using the data don't have to generate them themselves [10:54:04] and they wouldn't have to be stored in the database either, that would save a lot of space I bet [10:54:36] (but maybe there's a good reason it shouldn't work like that, I don't know much about wikimedia internal stuff, it's just my fantasies :D) [12:13:51] My current idea, and we'll all discuss this more before we start implementing it, is to have something like a new language code, "abstract" [12:14:28] And then use this for fallback, but not display the abstract content but use it to generate the respective natural language [12:15:21] So, not a bot who does that, but extend Wikidata itself to call the render function whenever it falls back to abstract [12:15:39] (plus tons of caching) [12:49:39] Descriptions can have functions. There are currently many NLP techniques that are based on glosses. (re @amire80: I mean, for what purpose. [12:49:40] What's the problem if an item doesn't have a description? [12:49:41] If a description can be generated by a bot, maybe it's not necessary in the first place?) [12:50:19] The interaction between Wikidata and Abstract Wikipedia can have many benefits. [12:52:04] Z-items can be used to extract semantic relations from raw texts based on machine learning techniques. [12:53:11] This is the point. (re @vrandecic: So, not a bot who does that, but extend Wikidata itself to call the render function whenever it falls back to abstract) [14:30:39] Cross-posting this to a bunch of channels, sorry about the spam :) [14:30:41] This Wednesday there will be a Language showcase meeting on the topic of module localization: [14:30:42] * https://www.mediawiki.org/wiki/Wikimedia_Language_engineering/Showcase [14:30:43] * https://meta.wikimedia.org/wiki/Meta_talk:Babylon#Language_showcase_about_module_translation [14:30:44] This may (or may not) be the first step to sharing templates and modules between the WMF projects, which is related to Abstract Wikipedia, so I think it's relevant enough to mention here :) [14:31:18] Of course, it is worth to be mentioned. (re @amire80: Cross-posting this to a bunch of channels, sorry about the spam :) [14:31:18] This Wednesday there will be a Language showcase meeting on the topic of module localization: [14:31:20] * https://www.mediawiki.org/wiki/Wikimedia_Language_engineering/Showcase [14:31:21] * https://meta.wikimedia.org/wiki/Meta_talk:Babylon#Language_showcase_about_module_translation [14:31:22] This may (or may not) be the first step to sharing templates and modules between the WMF projects, which is related to Abstract Wikipedia, so I think it's relevant enough to mention here :)) [14:32:15] This is interesting now. [17:37:52] Proposing this logo to Abstract Wikipedia [17:38:22] [[File:Wikidata logo proposal (Sonia).svg]] [17:42:10] not bad [17:58:13] I like it. [18:04:58] We can do a derivative with eight or ten edges [18:05:49] We can discuss this. [18:30:56] @vrandecic Where the Wikimedia Language Code has been put in this page [18:31:22] [[File:Wikilambda early mockup succession renderer english.png]] [19:23:24] I like it, but I am wary of using a Wikidata logo for AW. The defining feature of AW should not be a reliance on a knowledge graph. (Unless I am misunderstanding what the logo is intended to mean in the AW context ...) (re @Csisc1994: [[File:Wikidata logo proposal (Sonia).svg]]) [19:32:24] It meant to be a wiki of functions (re @Chris: I like it, but I am wary of using a Wikidata logo for AW. The defining feature of AW should not be a reliance on a knowledge graph. (Unless I am misunderstanding what the logo is intended to mean in the AW context ...)) [19:32:48] So, edges can be substituted by arrows and this will absolutely work. [19:36:27] So is the idea that red is wiki of functions, blue is abstract content, and green is generated natural language across supported languages? [19:37:58] Or are you just thinking of wikilambda, and blue and green are either part of the wiki of functions, or some generic application built on top of it? [19:38:45] I'm supporting this one (re @Chris: So is the idea that red is wiki of functions, blue is abstract content, and green is generated natural language across supported languages?) [19:49:44] +1 (re @Chris: So is the idea that red is wiki of functions, blue is abstract content, and green is generated natural language across supported languages?) [19:49:51] This is the idea [20:01:19] I like it (or something similar) then. Thanks for proposing it! [21:33:05] I like that logo too. [22:00:22] Not bad! I fear however that it has too much detail to be shrunk down (re @Csisc1994: Proposing this logo to Abstract Wikipedia) [22:01:12] https://tools-static.wmflabs.org/bridgebot/4d09b86b/file_448.jpg [22:06:13] what happens if the lines are removed? [22:06:43] The its just dots :) [22:06:52] The it is just dots :) [22:07:08] https://tools-static.wmflabs.org/bridgebot/5c693b97/file_450_tgs.webp [22:08:58] Then it is just dots :) [22:22:48] I think it kinda works : https://tools-static.wmflabs.org/bridgebot/cb2ac8ca/new_logo.png [22:23:04] Now that looks more fun than the original (re @moebeus: ) [22:23:10] rotfl (re @moebeus: ) [22:29:57] And a name proposal as well? (re @moebeus: )