[06:50:46] How do I get the "Generate" buttons back in Abstract Wikipedia after I’ve already generated the text? [06:50:47] I added a missing label in the required language on Wikidata, and I want to see the new text on the abstract article page instead of the "Q..." identifier. [06:50:48] https://abstract.wikipedia.org/view/ru/Q127 [08:16:30] The old cache will expire in about 24 hours, so it should work then. The team has recently been working on letting users reload the cache, but I don't think you can do it from AW yet. (re @alexander_mart: How do I get the "Generate" buttons back in Abstract Wikipedia after I’ve already generated the text? [08:16:30] I added a missing label i...) [12:55:32] I think most Julian dates from Wikidata are now supported in the common NLG functions. : https://tools-static.wmflabs.org/bridgebot/a8888ec0/file_83854.jpg [13:38:27] the meetng starts in ~20 minutes? [13:38:42] ah, no, it was 16:00 UTC. [13:38:47] ignore me! [14:01:51] I hope the meeting goes well. 🎃 [14:02:14] looking forward to it! [15:52:06] ah, it just tells me that it's not working in my browser. and I should download firefox. ... using firefox. [15:53:31] quick reminder that this is happening in 7 minutes, see you there! (re @Sannita: Hi all! A quick message to inform you that Denny will lead a quick first introduction to syntactic phrase functions in the next ...) [15:53:45] Works for me (Firefox 155.0.1 on Ubuntu) (re @wmtelegram_bot: ah, it just tells me that it's not working in my browser. and I should download firefox. ... using firefox.) [15:54:43] I honestly don't think it's my browser. my best guess is that something random is broken and that is their default error message. blame the user. [15:55:05] I will try using another ISP, let's see if that changes things. [15:58:04] yes, that was the case. if I come from a domesting and not a cooperate ip the very same browser is fine... [17:26:10] thanks again for the meeting/presentation. :) [17:26:34] thank you for joining! [17:27:00] the recording will be available in a couple of days, I need some time to convert it into a Commons-compatible format [17:27:35] * phschafft nods. [17:31:14] Have you tried https://video2commons.toolforge.org ? It converts the video server side and auto uploads it when done so you can kind of post-and-forget. (re @Sannita: the recording will be available in a couple of days, I need some time to convert it into a Commons-compatible format) [17:31:36] not yet, probably I'll give it a try (re @Jan_ainali: Have you tried https://video2commons.toolforge.org ? It converts the video server side and auto uploads it when done so you can ...) [17:31:45] I usually use Handbrake on my PC [18:03:01] still thinking a little bit about things that have been discussed today. [18:03:57] was just thinking about a function that returns the name for a thing once per sentence and a pronoun all the other times it is called within a context (sentence, paragraph). [18:06:31] 'f(Tim) won becase f(Tim) was good in the games f(Tim) played' -> 'Tim won becase he was good in the games he played'. while I suspect most/all languages to prefer name the first time, in theory it could also return it in any other position. with the contract that it returns it once, but where ever it's right as per language. [18:08:39] I tried it once, but it does take sometime. I recall one day or more. (re @Jan_ainali: Have you tried https://video2commons.toolforge.org ? It converts the video server side and auto uploads it when done so you can ...) [18:11:34] It, of course, depends on the size of your video. But the advantage is that you enter all the data and then let the server encode and upload rather than you needing to wait for it to encode on your computer first making you the final bottle neck. (re @jsamwrites: I tried it once, but it does take sometime. I recall one day or more.) [19:23:07] Yes… the limitation is that all your f(Tim)s have to be evaluated within the same function call, unless we have a mechanism for carrying state between evaluations. In a “traditional” NLG pipeline, these are called “referring expressions” and resolved as late as possible, as Ariel suggested. Conceptually, you can think of f(Tim) as a unique placeholder token: resolution [19:23:07] [19:23:08] can include inflection, hyperlinking and styling, as well as selection from among available alternatives. (re @wmtelegram_bot: 'f(Tim) won becase f(Tim) was good in the games f(Tim) played' -> 'Tim won becase he was good in the games he played...) [19:57:51] yes, there must be state scoped to the context level (I would argue that private to f() would be best so other functions would not mess with it). which is something the env needs to support in some way. [19:58:26] please bear with me for a bit. I'm only here for a few days now and the technical details are all still blurry to me. so I hover on the concept layer for a bit. [20:06:12] I don’t see why it would be private. It is at least conceptually feasible that every required referent could have its referring expressions pre-computed. The options would generally be the same across the whole article and in every other article that refers to the same referent. (re @wmtelegram_bot: yes, there must be state scoped to the context level (I would [20:06:13] argue [20:06:14] that private to f() would be best so other funct...) [20:11:45] the state being private. not the function. [21:37:12] I did not join the meeting, what's new with syntactic phrase functions? [21:41:24] I guess you and I wait the recording [21:45:49] Yes, it will definitely be worth watching.