[04:05:21] (03PS1) 10Catrope: SpecialORESModels: Sort filters in the expected order [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458630 [04:05:42] (03CR) 10Catrope: "This isn't the cleanest way to do this, but I'm also not sure if there even is a clean way to do this..." [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458630 (owner: 10Catrope) [07:40:35] 10Scoring-platform-team (Current), 10ORES, 10Patch-For-Review, 10User-Ladsgroup: Use poolcounter to limit number of connections to ores uwsgi - https://phabricator.wikimedia.org/T160692 (10akosiaris) >>! In T160692#3126539, @akosiaris wrote: > > I highly doubt there is a need for an ORES dedicated poolco... [07:45:30] 10Scoring-platform-team (Current), 10ORES, 10Patch-For-Review, 10User-Ladsgroup: Use poolcounter to limit number of connections to ores uwsgi - https://phabricator.wikimedia.org/T160692 (10akosiaris) [07:45:35] 10Scoring-platform-team (Current), 10ORES, 10Operations, 10Patch-For-Review: Spin up a new poolcounter node for ores - https://phabricator.wikimedia.org/T201824 (10akosiaris) 05Open>03Resolved a:03akosiaris Done in T203465 [08:11:12] (03CR) 10jenkins-bot: Localisation updates from https://translatewiki.net. [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458681 (owner: 10L10n-bot) [09:31:55] o/ [10:16:07] (03CR) 10Ladsgroup: [C: 032] "\o/" [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458630 (owner: 10Catrope) [10:23:28] (03Merged) 10jenkins-bot: SpecialORESModels: Sort filters in the expected order [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458630 (owner: 10Catrope) [10:26:25] (03CR) 10jenkins-bot: SpecialORESModels: Sort filters in the expected order [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458630 (owner: 10Catrope) [10:51:12] 10Scoring-platform-team (Current), 10MediaWiki-extensions-ORES, 10ORES-Support-Checklist, 10Patch-For-Review, 10User-Ladsgroup: Change mentions of wp10 to articlequality in products - https://phabricator.wikimedia.org/T203080 (10Ladsgroup) Yes, that's true but looking at the PageTriage patch it definitel... [11:31:24] since it's Friday I decided to do light-weight bug fixes [11:31:26] :D [11:40:48] 10Scoring-platform-team, 10ORES: ORES service gets stuck reporting "server overloaded" even after load returns to normal - https://phabricator.wikimedia.org/T175654 (10Ladsgroup) @awight you said this is done. Should we close this? [11:49:25] 10Scoring-platform-team, 10ORES, 10Wikimedia-Logstash, 10Patch-For-Review, 10TestMe: Send celery logs and events to logstash - https://phabricator.wikimedia.org/T169586 (10Ladsgroup) >>! In T169586#3604923, @awight wrote: > Looks like we already have messages going to logstash, per T175736#3604901. Conf... [12:28:31] 10Scoring-platform-team, 10ORES: Model information UI (graphs and statistics) - https://phabricator.wikimedia.org/T140364 (10Ladsgroup) We have the model representations in [[https://en.wikipedia.beta.wmflabs.org/wiki/Special:ORESModels | the extension now. ]], overall I think we should have a dedicated servic... [12:28:32] 10[1] 04https://meta.wikimedia.org/wiki/https://en.wikipedia.beta.wmflabs.org/wiki/Special:ORESModels [14:21:42] forgot to change back [14:22:00] * halfak looks for fun Friday work. [16:24:08] Changing location and I'll grab lunch from the U. So I'll be back on in 1.5 hours. [18:23:18] o/ [18:47:57] https://etherpad.wikimedia.org/p/jade_ores_research_ideas\ [18:47:59] https://etherpad.wikimedia.org/p/jade_ores_research_ideas [18:48:15] For potential interns and other people we might like to work with. [19:12:19] (03PS1) 10Umherirrender: Run stylelint also for less files [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458872 [19:14:51] (03CR) 10jerkins-bot: [V: 04-1] Run stylelint also for less files [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458872 (owner: 10Umherirrender) [19:16:32] halfak: so I hear you are wary of ORES-influenced judgments making their way into JADE [19:17:30] (03CR) 10Jforrester: "recheck" [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458872 (owner: 10Umherirrender) [19:20:09] (03CR) 10jerkins-bot: [V: 04-1] Run stylelint also for less files [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458872 (owner: 10Umherirrender) [20:13:41] harej, not wary exactly [20:13:57] More I just want to be able to differentiate them from "pure"er judgments [20:14:36] Sorry was AFK talking to research's newhire about working remotely. [20:14:45] He's visiting the university today. [20:17:24] I see we've expanded the Minneapolis contingent? [20:18:43] Yup :D Starting in Oct, there will be 6 of us! [20:19:38] you're going to need to start getting restaurant reservations [20:22:00] i talked about it with awight yesterday; we weren't sure how to force provenance information into JADE [20:22:18] my idea was to use edit tags, which shouldn't be a problem? [20:25:18] 10Scoring-platform-team, 10ORES, 10Documentation, 10User-srodlund: Feedback on ORES threshold optimization docs - https://phabricator.wikimedia.org/T203505 (10srodlund) [20:26:29] harej, I don't like that because it's not part of the content. [20:26:39] You need to process the revision history to get it. [20:27:26] So if you get a dump of the most recent version of the content, you don't know how those judgments were provided or who they were provided by. [20:27:45] I think the inherent subjective nature of judgment means that authorship and origin are extra important. [20:28:26] so I have an idea, but it requires never allowing raw JSON to be edited through the UI [20:29:53] (We don't need a sophisticated UI as such, just something basic that would let us achieve this goal.) [20:30:30] We have a provenance tag associated with each judgment. If you edit directly through the wiki, this field is left blank, and it would not be possible to manipulate it as it would not be exposed in the UI. Client applications have the opportunity to fill in this field via API, and would be expected to do so. [20:31:01] "provenance": "huggle-3" [20:31:03] Something like that. [20:32:20] Why would we not let people edit it? [20:33:02] The idea is that if you leave it blank it means you're a regular human client, but generally I want to avoid the situation where someone pretends to be Huggle to mess up the data. [20:33:16] (03PS2) 10Umherirrender: Run stylelint also for less files [extensions/ORES] - 10https://gerrit.wikimedia.org/r/458872 [20:35:22] I'm not sure we should design against abuse before the abuse happens. Can't we just have people police that? [20:35:31] Also what incentive does someone have to fabricate origin? [20:35:49] It seems that our biggest problem would be having people not know what to put into that field at all. [20:36:55] And I don't think it makes sense to expose a field that is confusing and not useful. [20:38:52] MediaWiki doesn't give us great options here. The originating client is fundamentally metadata and should be captured as such. If the concern is inconvenience to researchers, we can set up a data product that properly blends human-generated data and metadata together. [20:41:06] To answer your question, petty vandalism is definitely a thing that happens, and this doesn't seem like the best use of the honor system. [20:41:47] Having it be a hidden key would let the data be captured as part of the content. [20:41:58] But without exposing it to vandalism. [20:43:16] I don't think the context and individual making a judgment is secondary-metadata. I think it is primary-metadata. [20:43:29] The wiki doesn't rely on the honor system. people clean up vandalism. [20:46:35] I feel like "researchers" are being considered a secondary user here. [20:46:57] But I think "researchers" includes anyone who is using JADE data for auditing or any type of analytics. [20:48:36] So, when I visit a web page, do I get to arbitrarily define what my user agent is? [20:50:02] Does my user agent get exposed as a custom-editable field that can be edited after the fact in the logs? [20:53:29] As for researchers, my goal is to produce high quality data. But ultimately it's up to the contributors to do the legwork of producing this data, and they need to have the best tools for the job. [20:54:29] Hmmmm [20:54:35] I can see how a "context" field can work now. [20:54:47] For humans, they can describe what activity they were doing. [20:55:20] Still seems like a lot of extra work, but if we're exposing the field. [20:56:41] You do get to arbitrarily define what your user-agent is. [20:56:59] It's not editable after the fact in the logs (unless you're the sysadmin) [20:57:12] But signatures are editable on talk pages and we police that just fine. [20:57:46] harej, I expect that contribution interfaces would auto-fill the field in 99% of cases. [20:57:59] But if someone wants to fill it in themselves manually, cool. [20:59:06] If someone manually edits a judgment, what do they fill in? What if someone creates a judgment with Huggle then someone else changes it? [21:02:01] If someone manually edits a judgment, they'll be manually applying the spec. They might put nothing or type "manual" [21:02:15] If someone else changes it, then I guess that'd be reverted. [21:02:28] In the same way that editing someone else's signature is reverted. [21:02:40] What if judgments are collaboratively edited? [21:02:44] --Abraham Lincoln [21:02:55] Right. [21:03:02] Like wiki pages or talk pages. [21:03:18] In wikibase, you can do these things too with evil intention! [21:03:23] Yet no one does it. :) [21:03:30] More like content pages, where changing absolutely anything is fair game, as opposed to talk pages, where the honor system holds. [21:03:31] See "update item" in the API [21:04:00] I'm thinking of a scenario where we have one defined judgment for a given {diff, revision, page}. People will disagree, at which point the provenance becomes fuzzy. [21:08:20] Oh? [21:09:35] If we allow multiple judgments on a given JADE page, then yes, everything you say holds. [21:11:00] Oh! yes. You're right. [21:11:30] If it's just one judgment and consensus-based, then I agree that our best option is change tags and edit comment hash tags,. [21:15:13] Basically I've come around to your original point of view, which is that judgments are collaborative, and this is the differentiator between JADE and an arbitrary Postgresql instance [21:15:34] And we should embrace it for all its interesting ugliness. [21:18:09] \o/ Now the question of what kind of ugly. [21:18:32] I'm down for the complex version where people get tied to their opinions from an origin/context. [21:18:45] Or the one where it is just a judgment and BRD applies. [21:19:03] BRD will apply anyway, but filing disagreement only works in a structured way in one case. [21:20:31] So, I'm glad you agree that "JADE as a complement to ORES" isn't a red herring. I think of JADE like that, but I am also currently working on giving an identity to JADE that is independent of ORES. [21:20:43] JADE is its own thing, in addition to being related to ORES. [21:20:45] +1 for that. [21:21:12] One of the ideas I have been pushing is that JADE is a collaborative judgment repos and only some judgments are worth automating. [21:21:24] E.g. you might imagine arbcom or AFD moving to JADE :| [21:21:39] Not sure what I think of that but it's an interesting example. [21:22:36] It's theoretically possible but we'd need JADE to do a lot more things before we could consider that. [21:22:39] Let's not get ahead of ourselves! :D [21:24:53] My idea is to refer to JADE as a "collaborative annotation system," an extensible system for annotating all sorts of things. [21:29:04] I think that's a good way to explain it. [21:29:17] One thing I was considering was re-branding wiki labels. [21:29:37] To what? [21:29:44] Wiki labels is a human-computation system -- not really a general wiki-style collaborative labeling system. [21:29:54] JADE is really "Wiki labels" [21:30:09] Where the current wikilabels is really all about batches. [21:30:14] or "campaigns" [21:32:51] Anyway, that's really where the thought ends. [21:33:03] I'm not sure I'm ready to suggest we re-name JADE to wiki-labels. [21:33:18] But your statement about the "anything annotator" made me remember that. [21:33:30] If you're open to renaming JADE my idea was to call it something really drab like "Annotations" [21:33:36] (not "Annotations" but something like it) [21:34:01] Ugh. [21:34:12] We're bikeshedding. [21:34:21] And I don't like that color :P [21:57:58] OK I'm out of here. Have a good evening (or whatever), folks. [21:58:03] o/ [23:04:02] 10Scoring-platform-team, 10JADE, 10drafttopic-modeling: Discuss whether we can deploy JADE for draft topic without fixed categories - https://phabricator.wikimedia.org/T203032 (10Harej) p:05Triage>03Low [23:04:19] 10Scoring-platform-team (Current), 10JADE, 10Patch-For-Review: Use local articlequality assessment scale for JADE. - https://phabricator.wikimedia.org/T203030 (10Harej) p:05Triage>03Low [23:05:18] 10Scoring-platform-team, 10JADE, 10drafttopic-modeling: Discuss whether we can deploy JADE for draft topic without fixed categories - https://phabricator.wikimedia.org/T203032 (10Harej) I think we can defer on articlequality for the time being. Let's get JADE out the door first, then we can worry about this. [23:27:30] (03PS1) 10Reedy: Minor cleanup [extensions/JADE] - 10https://gerrit.wikimedia.org/r/458931