[00:30:22] That's true. I've created [[abstract:Project:Manual of Style]] as a draft, feel free to improve! One thing I eventually want to do is split it into a manual of style for function authors and a manual of style for content authors. (re @u99of9: For individual languages, as far as it's possible we'll probably want to imitate/follow the appropriate Wikipedia MOS where [00:30:23] avai...) [00:43:51] I just had a go at Z35081, but it looks like it has one of those unresolved reference issues. (re @u99of9: I'm stoked to see references in play, that's a game-changer. Thanks David .) [00:48:10] Is there any way that the community can change the suggested fragment types? I thought there was but I forgot it. [00:50:09] T425895 (re @u99of9: I just had a go at Z35081, but it looks like it has one of those unresolved reference issues.) [01:18:20] Oops, nevermind, I was comparing a reference to a single element list of references. Now working. (re @u99of9: T425895) [02:45:51] Feel free to expand Z35087, I expect it will be very highly used (called from Z35085) in AW. [05:31:16] The point that I was trying to make is that there are a lot of more important reasons why it is necessary to keep a shared context. For many languages (perhaps for all languages) it is necessary in order to produce grammatical output (re @u99of9: Yes, interesting example. I think there are ways this could be dealt with other than trying to hold full article [05:31:16] context. One mi...) [05:34:09] But is it the case that these contextual factors *can't* be flagged by human editors? Or do you judge that it will become too cumbersome? [05:36:41] In English I doubt it is necessary, but without some kind of contexts you end up with very boring repetitive language. [05:39:04] Agree. Do we even need to repeat those or can we just refer to them? (re @u99of9: For individual languages, as far as it's possible we'll probably want to imitate/follow the appropriate Wikipedia MOS where avai...) [05:58:29] A very important context factor is whether the concept has already been mentioned, in which grammatical context... those are impossible to manually encode, since they are language-specific (re @u99of9: But is it the case that these contextual factors *can't* be flagged by human editors? Or do you judge that it will become too cu...) [06:02:29] I remember you that without such shared context, the use of pronouns is impossible. And also the handling of definitiveness too: theoretically, even an article with 2 fragments needs a shared context, if you don't want it to look like a list of Wikidata claims (re @u99of9: But is it the case that these contextual factors *can't* be flagged by human editors? Or [06:02:29] do you judge that i [06:02:30] t will become too cu...) [08:28:30] But it is still very fundamental, and this is why I think that solving this problem should be a priority (re @u99of9: Although I agree that theoretically it would be lovely to know the full context of every fragment, I fear that waiting for or re...) [09:53:25] At the very least, we should consider whether the basic content functions deliver their context encoded within the HTML or (as a standard) by a content+context pair. I have always assumed it should be the former (see these comments, (https [09:53:25] //meta.wikimedia.org/wiki/Talk:Abstract_Wikipedia/Archive_2#c-GrounderUK-2020-07-19T23:11:00.000Z-DVrandecic_(WMF)-2020-07-14T05:15:00.000Z) [09:53:26] for example), but I’m open to persuasion. [09:53:27] I’m less convinced by the idea of article-level context because the most salient linguistic context is usually what has just been stated or what is about to be stated. To me, this implies an inter-call context, which is difficult to achieve in a pure-function approach, because it ultimately depends on both preceding and subsequent function calls. [09:53:29] However, if contextual information is embedded within the HTML itself, it becomes conceptually tractable for a final article or section render to consider both earlier and later results when refining the previously generated text (and, if appropriate, the enclosing HTML). [09:53:30] In other words, the per-language HTML itself is the required context. And this is true from the moment the first call is inserted, since even the result of that call is affected by the page title. (re @dvd_ccc27919: But it is still very fundamental, and this is why I think that solving this problem should be a priority) [10:12:35] I am not just speaking about minor refinements, I am speaking also about entire different sentence structures depending on the context. This means that the rendering functions themselves need to have access to the entire context, not only by a final refiner function (re @Al: At the very least, we should consider whether the basic content functions deliver their [10:12:35] context encoded wi [10:12:36] thin the HTML or (as a...) [10:20:17] I know. But I didn’t say anything about “minor”. And, as I said, even the result of the first function call depends on the entire context at the time it is evaluated to its initial “final” form. (And the next function call that is inserted depends on that result, and then the subsequent “final” form depends on both results… and so on). (re @dvd_ccc27919: I am not message> [10:20:18] just speaking about minor refinements, I am speaking also about entire different sentence structures depending on the c...) [10:25:11] And that's why, if we hadn't time execution constraints, we should generate the entire article in a single function call (re @Al: I know. But I didn’t say anything about “minor”. And, as I said, even the result of the first function call depends on the entir...) [10:27:10] I don’t agree that that is a logical conclusion or a practical approach, even ignoring the performance. (re @dvd_ccc27919: And that's why, if we hadn't time execution constraints, we should generate the entire article in a single function call) [10:27:40] What are the alternatives? (re @Al: I don’t agree that that is a logical conclusion or a practical approach, even ignoring the performance.) [10:31:04] What I just described: each function call is evaluated in the context of the results of all existing function calls, just as each conventional article is amended in the context of the whole article as it currently stands. [10:33:28] For an example of which sort of context we need, see https://www.wikifunctions.org/wiki/Wikifunctions:Type_proposals/Semantic_unit/Douglas_Adams [10:33:29] We need to uniquely reference to real life objects, and we need context on when and how those objects have been alrealy mentioned. For example, we might need to speak about a single computer A, that is part of a set B of three computers. We cannot just rely on QIDs, since there is no QID to refer to the single computer A we are talking about, or the set B (re@Al: [10:33:29] At the very lea [10:33:30] st, we should consider whether the basic content functions deliver their context encoded within the HTML or (as a...) [10:33:52] How is it possible? (re @Al: What I just described: each function call is evaluated in the context of the results of all existing function calls, just as eac...) [11:20:36] The challenge, it seems to me, is deciding which HTML attributes are implied. Clearly “computer A” is some instance of the class identified by the item id, and its particularisation should (in an editorial sense) be clear from the prior content, even though it is transient (“the” computer in context at the point the function call is executed). If computer B is introduced [11:20:36] [11:20:38] (editorially) it may need to be clarified that this is not the same instance. Whether it is the same or not depends on where the function call is positioned, but at any point in time, the call either introduces a new instance or it refers to a previous instance. [11:20:39] Here we get into the complexities of user-interface design, but it is conceptually straightforward to assert that “this” instance of an item is the same as the explicit item in some other function call (say, the one “–3” function calls prior to the current one). Whether the generated HTML should always include a reference attribute is an open question, because some item [11:20:39] [11:20:41] s are contextually unique, but perhaps it is simpler to assume that it does. Then the question becomes even more one of how the UI exposes such references and allows their specification within a new function call. But I would expect such a requirement to be relatively uncommon, just as I would rarely insert something like “call it computer A” into conventional text (such [11:20:41] expl [11:20:42] icit referring expressions are likely to prove challenging in abstract content, but we cannot solve every problem immediately). (re @dvd_ccc27919: How is it possible?) [11:51:42] I don't see how this approach would be better than the one proposed by @mahir256 , which is simply to include (in the initially-provided "context") fake Wikidata items (or something conceptually analogous) for every real life object/collection of objects that do not already have an identifying real QID, and then refer to them in the abstract content (re @Al: The [11:51:42] challenge, it see [11:51:42] ms to me, is deciding which HTML attributes are implied. Clearly “computer A” is some instance of the class...) [12:13:23] I think the only material difference is that I assume no prior context other than that which can be extracted from the HTML, so it is generally implicit. The participants in a single function call still need to be particularised if they are not contextually unique (which is what I mean by “introduced”). But there are no “fake” items, only references to content that exists [12:13:23] [12:13:24] elsewhere in the article, like “in the previous sentence”, which is a textual anchor that functions as a reference (that is, it is one way to render the “–3” in my previous reply, which, in practice, would be some HTML attribute). (re @dvd_ccc27919: I don't see how this approach would be better than the one proposed by @mahir256 , which is simply to include (in the ini [12:13:24] [12:13:26] tially-...) [12:21:14] The entire discourse is a bit confusing; when I have time I'm moving it on Wiki in a more organized way and with explicative diagrams [14:17:34] The corner is in 3 hours I think [14:18:32] The link to participate is: https://meet.google.com/xuy-njxh-rkw [14:19:17] Hi all! Our next *Volunteers' Corner* will be held today at 17:30 UTC. (https://zonestamp.toolforge.org/1778520600) [14:19:18] If you have questions or ideas to discuss, or you want to get in touch with the dev team, please join us! The link to the meeting is https://meet.google.com/xuy-njxh-rkw [14:19:20] Hope to see you there! [14:19:21] (re @Sannita: Hi all! Our next Volunteers' Corner will be held today at 17:30 UTC. [14:19:23] If you have questions or ideas to discuss, or you want to ...) [14:19:35] (sorry, I forgot to make the message last Friday [14:36:17] The explanation are on [[abstract:User:Dv103/Abstract articles architectures]]. Al I've not inserted your proposal because I don't understant what you meant, so feel free to add it yourself (re @dvd_ccc27919: The entire discourse is a bit confusing; when I have time I'm moving it on Wiki in a more organized way and with explicative dia...) [14:36:47] This time I'll try to participate (re @Sannita: Hi all! Our next Volunteers' Corner will be held today at 17:30 UTC. [14:36:48] If you have questions or ideas to discuss, or you want to ...) [16:08:06] Is there any way to get the statement value from a property (rather than from an item)? [16:08:08] Like from P178 (developer), I could specify P6104 (maintained by WikiProject) and get the two value items as a typed list [16:08:09] similar to all those "get statement from Wikidata property for item" but the item itself is a property, is what I was aiming for [16:14:47] https://www.wikifunctions.org/wiki/Wikifunctions:Catalogue/Wikidata_operations/Property [16:14:48] It seems like there are not a lot of functions dealing with Properties yet, but they can still be created (re @indiearblast: Is there any way to get the statement value from a property (rather than from an item)? [16:14:50] Like from P178 (developer), I could spe...) [17:03:29] This is happening in 30 minutes, see you there! (re @Sannita: Hi all! Our next Volunteers' Corner will be held today at 17:30 UTC. [17:03:30] If you have questions or ideas to discuss, or you want to ...) [21:07:34] I have added two tests for Z832. Could someone with required user rights connect them? [21:13:14] And for Z862 as well please. [23:05:48] This is something only staff can do I believe (re @Ameisenigel: I have added two tests for Z832. Could someone with required user rights connect them?)