[00:00:15] so, ways forward that we agree on, I think: [00:00:46] 1. use the 12000+ questions in the database somehow, for self-serve help [00:01:20] 2. get better at inviting more of the right people, fewer of the "less right" people (or maybe just more of the right people), and invite them at the right time [00:02:11] oh, and I already have the diffs for all the answers, and what question they're answering, in a table. [00:02:49] It would be good to talk to manybubbles about making the Teahouse searchable. [00:03:11] It's possible that he'll say "Meh, let me just flip this config variable." [00:03:45] If we can make it searchable (maybe a search API call?), we can start experimenting with users. [00:03:54] MWahahaha [00:04:50] Also, 3. talk to Teahouse hosts -- probably informally at first and formally later. [00:04:51] I mean, it's kind of searchable now. There's an archive. [00:05:01] linky? [00:05:16] https://etherpad.wikimedia.org/p/DRMethodsMenuSite [00:05:20] um, no [00:05:23] hold on [00:05:29] https://en.wikipedia.org/wiki/Wikipedia:Teahouse/Questions [00:05:34] archive box on the right. [00:05:46] 403 archives and counting [00:06:08] https://en.wikipedia.org/wiki/Special:Search?search=why+was+my+article+deleted&prefix=Wikipedia%3ATeahouse%2FQuestions%2FArchive&fulltext=Go+find+it%21&fulltext=Search [00:08:14] even the number of results says rejection :( [00:09:00] Emufarmers: yep :( [00:09:22] re: talking to Teahouse hosts, a first step might be asking them to share the heuristics they use to triage requests, and craft their responses to particular kinds of requests [00:10:37] J-Mo, yeah. Not the kind of search results I was hoping for, but maybe we can work with this. [00:11:58] re: FAQ, halfak [00:12:44] one of the issues with the way Wikipedia handles newcomer onboarding is that "page full of all the links" approach, that is so intimidating and overwhelming [00:13:23] but if we can extract links from answers to questions related to a user's question, we could potentially suggest a much smaller set of Help/Tutorial/Policy pages to them. [00:13:31] smaller and more relevant [00:25:41] I still think buy that a message tailored to the current situation will be best. [00:25:49] And it is best if that message is delivered by a human. [00:26:35] I agree completely. [00:26:53] one final thought, halfak, and then I'm going to make dinner. [00:27:08] Kind of a half baked one ATM, but interested in hearing your thoughts [00:27:35] one potential issue (and I think you alluded to this already) is that as we refine our ability to target "good faith newcomers" and send them invites, we may start to really reify a "have's" vs "have not's" scenario. Right now, the invite criteria are really general, and not obviously exclusionary. The current limiting factor is the number of invites I'm allowed to send (per my BotReq), and to a lesser extent the number of questions hosts can [00:29:06] algorithm and algorithmic practice are amoral, but we can be negligent by applying them to real, live people and not checking to see if they are causing harm. [00:29:32] If we're signing on to experiment with people's work patterns and emotions, we have a moral obligation to find out if we are systematically excluding some people. [00:29:44] I agree [00:29:50] This is a major focus of work on the revscoring project right now. [00:30:08] The best part is that our academic area totally agrees and would love to see such work published :D [00:30:25] MWAHAHA give me ALL the citations! [00:30:32] Indeed. [00:31:02] It's nice to find a healthy set of motivators in the mess of crap metrics. [00:31:21] indeed indeed. [00:31:41] alright, my brain is shot. food now. more conspiring anon! [00:31:46] Have a good night. More thinky-talky later :D [00:31:54] 'night [00:45:36] halfak: btw we are trying to replicate elasticsearch indices to labs [00:45:45] that'll allow more complex arbitrary queries from users [00:45:56] and are collecting use cases to justify possible hardware budget [00:46:20] halfak: https://phabricator.wikimedia.org/T109715 has details and a place for you to register support [06:12:23] https://meta.wikimedia.org/w/index.php?title=It.m.wikipedia.org&diff=14263871&oldid=11737032 [13:37:17] Thanks YuviPanda. Post made. [14:03:44] Nemo_bis, interesting! [14:04:05] We had 4 months of difficult anon editing. [14:05:10] Could that account for https://meta.wikimedia.org/wiki/Research:Active_editor_spike_2015_(July_update)? [14:05:24] It looks like no [14:05:25] But maybe [14:44:24] halfak: probably unrelated, but for sure it's quite easy to cause big fluctuations in numbers for registrations and 1+ edits contributors. [14:45:55] halfak: it's possible that by enabling the edit buttons on mobile we tricked some formerly active editor into clicking and reactivated their addiction, gaining a very active editor... but just wild speculation. [14:46:30] Though maybe we could at least measure whether the "additional" very active editors are reactivations or users who are very active for the first time. [15:27:18] the thing is, i tend to not be logged in on my phone [15:27:25] and logging in on your phone is something you do intentionally anyway [15:27:31] so that doesn't make it seem so accidental [15:38:13] good morning guillom J-Mo halfak Ironholds! o/ [15:38:29] hullooooo bearloga! [15:41:03] and good morning lzia!! [15:44:38] hi everyone [15:44:52] i'm currently adding symptoms to industrial chemical exposure to wikidata as part of my day job [15:44:58] the addition is taking place over 11,012 edits [15:45:08] good morning harej! [15:45:18] o/ bearloga and everyone [15:46:28] good morning bearloga and friends. [15:46:55] had any exposure to bromine? https://www.wikidata.org/wiki/Q21173316 [15:47:24] o/ [15:49:21] starting work on a forecasting project. stuck on the hardest part: naming the repo. considering: "cosa nonstradamus" but I guess I should save it until I'm building predictive models for organized crime activity [15:52:05] Which is hopefully your next project, right? [15:52:47] (Also see [[The Player (2015 TV series)]] [15:52:51] ) [15:54:11] guillom: real talk: the idea of using statistical inference to help detectives solve crimes was one of my motivations in college [15:55:20] Cool! [15:57:49] what other famous prediction makers throughout history can I make puns out of? [15:59:39] oh no [15:59:41] oh no no [16:00:54] somebody already made a module for python called pythia but it's not for making predictions/forecasting. it's "an extensible, object-oriented framework for specifying and staging complex, multi-physics simulations" [16:01:00] what a waste [16:23:08] milimetric: hey, around? [16:23:23] hey Amir1, yes [16:23:31] i have a few minutes before a meeting [16:23:35] you wanna get setup locally? [16:23:48] Okay, If you have time yes [16:24:06] I don't know if it's possible in few minutes (p.s. please check the PR) [16:24:08] so you forked / cloned, I see that, did you bower install/ [16:24:15] * milimetric looks [16:25:08] k, cool, no more translate class [16:26:13] I never worked with bower, [16:26:17] ah, ok [16:26:17] Installing it [16:28:07] milimetric: so I install trans-rec api using bower? [16:28:12] Amir1: just noticed a problem, the bower.json for some reason got excluded from the repo :/ [16:28:34] I'll need to fix that first, one sec [16:34:05] ok Amir1, I pushed the bower.json to master [16:34:20] if you rebase on top, you'll have api/static/bower.json [16:34:35] and so then you can do: "cd api/static" and then "bower install" [16:34:49] okay :) [16:35:26] that'll create api/static/bower_components which will have all the dependencies that are used as you see here: https://github.com/ewulczyn/translation-recs-app/blob/master/api/templates/index.html#L23 [16:38:07] milimetric: yeah [16:38:13] I see jquery and things like that [16:38:24] all installed [16:38:30] k, cool, I'm trying to figure out what the heck keeps happening to the sample data directory [16:38:38] I keep adding it and I think ellery keeps deleting it :) [16:38:40] one sec [16:42:43] watch the award ceremony: http://www.rtve.es/noticias/premios-princesa-asturias/directo/ :-) [16:43:59] ok Amir1, I pushed the test data and the updated readme [16:44:10] oh, Amir1 is here! :-) [16:44:15] hi milimetric, and Amir1. [16:44:17] specifically, check out https://github.com/ewulczyn/translation-recs-app/blob/master/api/README.md [16:44:20] hi leila [16:44:25] leila: hey :) [16:44:35] once you pull that, all you should need to do is: [16:44:35] python api/recommender_api.py --translation_directions test_translation_directions.json [16:44:49] and then browse to localhost:5000, the flask should be up there with some fake data loaded [16:44:49] we are working on making the app working for us [16:44:58] milimetric: thanks :) [16:45:03] it is great [16:45:07] I'm installing things [16:45:12] like pandas [16:45:16] :) [16:45:19] right, that's needed too, sorry [16:45:33] won't take a long time [16:45:38] until I rebase my own fork [16:45:56] leila: what's up there? [16:46:05] Amir1: health. ;-) [16:46:11] :D [16:46:15] Amir1: at the moment I'm watching the award ceremony. [16:46:29] Is it in Spanish? [16:46:42] Amir1: then, I want to talk to you to see what you're doing. Stewards are in town, meeting with them, and more. [16:46:58] yeah, Amir1. there should be an English subtitle I haven't found, yet. :D [16:47:11] great :) [16:47:41] I will definitely talk to you but I will have a meeting with Aaron in ten minutes our weekly session about revscoring [16:47:55] Amir1: let's have a chat about the app when you have time, maybe tomorrow? I'd like to get to speed with what you're working on and see how we can join efforts. [16:48:19] yeah [16:48:25] in Ashura [16:48:26] :) [16:48:47] just bear the noise and it would be okay :D [16:48:59] even at middle of AmirKabir [16:50:12] leila: Let's talk about the recommander system in tomorrow this time. OK ? [16:50:56] Amir1: tomorrow this time may be a bit late for me, probably 2 hours earlier. I'll send you an invitation. If we can't talk then, we can do it on Sunday. [16:51:18] Two hours sooner is okay with me :) [16:52:00] *for me [16:55:35] milimetric: It's working now for me :) [16:55:51] awesome, so now you can see your change and fiddle with it [16:56:08] maybe run it by leila, and then we can merge your PR [16:57:04] I want to change place of the button [16:57:21] from left it's first. I think it should be the second [17:00:55] * halfak watches DarTar wikignoming his pages :) [17:01:07] * halfak misses IRC DarTar [17:40:47] bearloga, want some homework? ;p [17:42:59] Ironholds: you know you can just be like "hey, would you mind doing ____?" because who in the history of the world who has gone to school at some point in their life has ever said "yes, oh my god, please give me some homework I really want some homework"? [17:43:45] bearloga, it's a paper I thought you'd find interesting! [17:43:46] Ironholds: but sure :P [17:43:47] not actual homework ;p [17:43:48] http://www.jting.net/pubs/2007/ting-ICRA2007.pdf [17:43:53] the title screams Mikhail [17:44:10] OOOOOOOOH [17:44:15] IT DOOOOOOEEEEESSSS [17:44:16] see? [17:44:32] it also screams "there should be an R package if it works" and "we should use it if it works" [17:44:39] it's pretty screamy [17:44:48] * harej is still thinking about structured image metadata and machine identification of images [17:44:52] there's a lot of screaming going on [17:45:08] Ironholds: thank you thank you thank you for this paper [17:45:26] no problem! [17:45:58] * bearloga is giddy like when he was a boy and the delivery man showed up with a package containing a new harry potter book [17:46:04] uh-oh [17:46:25] Ironholds: "uh-oh"? [17:46:36] bearloga, looks like the tryCatch in the automater isn't actually making it perpetuate [17:46:53] frack [17:47:05] it's the protonk [17:47:08] hey Protonk [17:47:21] bearloga, mind debugging? I'm sorry, I know it's meant to be your 20% day, but prod issues... :/ [17:47:40] Ironholds: sure no problem [17:47:47] thank you! I'm sorry! [17:47:52] hey hey [17:48:05] I'm just here for the username rules, lol [17:48:14] well, I'm just about to be in wikitech for that [17:48:15] what're the username rules [17:48:24] GOOD QUESTION [17:48:55] looking for where in the codebase you store the rules for unacceptable characters/sequences for usernames [17:48:56] hey Protonk [17:49:06] to see if I can add some client side validation to the signup page [17:49:16] isn't part of that governed by abusefilter as much as it is by mediawiki rules? [17:49:17] is this about colons? [17:49:26] or are you only looking strictly for technical violations? [17:49:27] partially, yes. :) [17:50:00] I'm looking for characters that will get stripped (colons, in this case, or at least the prefix WP:, haven't tested further) and chars that are disallowed [17:50:53] I'd also like to see if I can make calls to check for violations and/or similarity to let new users know name X isn't going to work before they hit submit [18:01:17] harej: I'm actually not sure how it's governed. TBH how all the moving parts interact on MW and on the specific instance of MW I care most about is pretty complicated and hard to grok w/o a little push. I'm getting crickets in #mediawiki so I think I'll just go see what's up from the docs (sigh…mediawiki docs) [18:02:26] lzia: madhuvishy we're getting 32G RAM instances for the art-recs project! [18:02:28] stand by... [18:02:34] yayyy [18:02:41] YuviPanda : yayyyy. :-) [18:02:44] YuviPanda: I might work on it today [18:02:52] this has been puppet week [18:02:55] heh [18:02:55] madhuvishy: :-) [18:02:57] nice [18:03:23] I can't figure out one bit with hiera though, i'll poke you [18:13:22] harej: #mediawiki came through. Yay, MatmaRex [19:47:43] Args! How come no one from my team is here for me to complain at about meetings that get moved minutes before they start! [19:52:08] halfak, I mean, you can complain to me, I'll just lack context [19:52:38] Na. It's good. [19:52:49] I'll just complain that they don't come on IRC. [19:52:51] halfak: call a meeting to complaiiin [19:52:58] Oooh good idea. [19:53:04] also remember IRCCloud accounts are still available for freee to all WMF employeeessss [19:53:04] And then I'll move it 5 minutes before it starts [19:53:14] YuviPanda, I did not know that. [19:53:18] How do I get free? [19:53:29] halfak: oh you ask me or OIT [19:54:14] Can haz? [19:54:29] halfak: sure [19:54:52] halfak: what's your WMF email? [19:55:12] ahalfaker@wikimedia.org [19:55:20] I don't have that account registered yet [19:55:23] I have one with my personal [19:55:34] so I've been told to send invites only to wikimedia.org accounts [19:55:47] you can sign up with that, I think? not sure if you can transfer it to your personal account [19:55:47] OK no worries. Just so long as it not being registered is OK [19:55:55] That's just fine [19:56:10] It's donor money. Doesn't have to be on my personal account. [19:56:31] halfak: invite sent [19:56:37] halfak: it's also not donor money. we got a 100 accounts for free [19:57:01] WOoo [19:57:06] Horray for goodwill [19:57:09] indeed [19:57:22] I've stopped using it though [19:57:35] * YuviPanda has been in soul searching mode for a while now [19:58:21] so no mobile IRC [19:58:31] Why did you stop? [19:59:32] yo [19:59:38] no [19:59:40] k [20:01:36] halfak: it was just taking up too much of my mental energy and whatnot [20:01:48] Oh. Just being online all the time? [20:01:57] also my computer is a core M and fairly underpowered and the IRCCloud redesign made it a lot slower [20:02:02] so I switched to weechat + glowing-bear [20:02:08] halfak: yeah and also being online on phone [20:02:10] Get new computer! [20:02:18] nah, I specifically got this a few months ago [20:02:31] it's super cheap, lotsa battery life, and runs vim just fine [20:02:40] I also have a chrome extension that auto closes tabs if I've more than 10 open [20:02:47] ha. [20:02:55] I have two browsers with three tabs each right now. [20:03:01] I'm very disciplined with my tabs. [20:04:02] I believe that I should offload discipline to external systems as much as possible, since even the tiniest sliver of discplilne takes an enormous mental effort... [20:05:37] YuviPanda, different school of thought, but I see discipline as a simplifying strategy that reduces mental load. [20:05:47] halfak: oh I totally agree. [20:06:03] halfak: I just don't have the capacity for it, so I spend a few hours building systems to take load off me. [20:06:21] the tab closer extension for example [20:06:31] enforces discipline in that I no longer have 400 tabs that feel overwhelming [20:06:36] But really, I think discipline is its own end. Squash discipline helped me get through some tough times in grad school. [20:06:39] and I finish purchases / fill forms in one go instead of flittering around [20:06:39] Good point. [20:06:55] since I could no longer be like 'let me go read that for a while since this form is too long' [20:07:02] because it might get closed! [20:07:06] Ahh! I see. [20:07:24] Yeah. That's still discipline -- you just have a mechanical coach! [20:07:33] indeed. [20:07:46] I too agree that discipline is necessary, and I just want to automate as much of it as possible [20:07:50] so I don't have to think about it [20:08:12] got diagnosed with ADHD a few years ago, and since I'm not on meds anymore I've to invent other ways. [20:08:18] sometimes fun, sometimes infuriating. [20:10:49] Gotcha. It's good to engineer your behavior. That's a meta discipline. [20:11:16] yeah [20:11:27] I'm working on a small script that'll allow me to arbitrarly block network connections [20:11:34] with predefined setups [20:11:46] so I'd hit a button, and get 25 mins of a particular network configuration [20:16:23] What sort of configurations would you preset? [20:17:58] in general, you should offload any and all executive functions to technology [20:18:10] scheduling, task management, none of that should be in your head [20:18:14] halfak: still tweaking. at the least, blocking off facebook, twitter, etc. [20:18:27] halfak: I'm also considering ways to limit IRC but not sure if that's feasible [20:18:30] harej, meh. I use my head for *a lot of stuff* [20:18:43] halfak: like, I can respond to pings but not just idle and backscroll [20:18:46] well congratulations! your executive function is better than mine [20:18:53] halfak: what harej said [20:18:55] YuviPanda, just ignore backscroll? [20:19:01] Ha. [20:19:07] halfak: no, the idea is to do this in 25min + 10min increments [20:19:15] halfak: so I'd turn this on and get 25mins of time when I can train myself to focus [20:19:27] so during those 25mins i'd like to ignore non direct pings [20:19:32] Gotcha. [20:19:46] Do you get notifications for any message in certain channels? [20:19:49] no [20:19:58] but I find it hard to just not make all the unread message numbers say 0 [20:20:15] so I end up tabbing to it every few minutes and reading allll the things [20:20:40] there are certain channels where i don't really care about the backscroll so i consistently ignore it [20:20:41] Gotcha. [20:21:12] it's more of a 'be aware of your limitations and work around them' than anything else. [20:21:18] How do I file YuviRequests in a good way for YuviBehavioralPatterns? [20:21:26] halfak: pings are totally ok! [20:21:34] and so are emails [20:21:57] YuviPanda, will you look at https://github.com/wiki-ai/ores-wikimedia-config/pull/39? [20:21:57] :D [20:22:04] sure [20:22:42] halfak: merged [20:24:14] Looked OK -- how I put it together? [20:24:30] Wasn't sure if you thought it should live in ORES, but the logging system was nice to configure from there. [20:24:33] And the output looks OK [20:25:49] Oh well. Thanks YuviPanda :) [20:26:15] halfak: I did look through it and it sounded sane and would do the use cases I'd want it to do eventually (send things to logstash) [20:26:29] Cool :) [20:26:44] BTW, did you see the bug I filed re. celery queue maxsize? [20:27:10] Was really hoping they'd turn around and say "Just do and it'll work" [20:27:11] https://github.com/celery/celery/issues/2874 [20:27:17] Might have to fix this upstream. [20:27:20] :( [20:27:47] halfak: we can kindof 'fix' it in redis by monitoring size of the key they use as a queue [20:29:21] YuviPanda, yeah, but that's super gross since the way that celery uses redis might change. If we figure out how to do it outside of celery, we can probably implement the fix within celery's redis adapter. [20:29:29] But yeah, gotta see the code first [20:29:32] indeed [21:35:50] Running away to do weekend stuff. Ping me on gchat if you need me. Ande have a good weekend! [21:35:51] o/