[08:36:58] <wikibugs>	 10Quarry: Cache user.get_user() - https://phabricator.wikimedia.org/T197551#4295691 (10zhuyifei1999) The user object isn't a serializable (or are we using pickle and it is serializable?). Anything in session has to be serialized in order to be stored into redis.  Though, filtering by id is a `eq_ref` kind of que...
[17:38:20] <psinghal20>	 Hi, I am Pratyush Singhal, a Google summer of code intern in wikimedia. I am working on the Article Finder tool(https://phabricator.wikimedia.org/T111416) for the WikiEducation Dashboard(https://github.com/WikiEducationFoundation/WikiEduDashboard). We were looking into the missing sections API(https://recommend-alpha.wmflabs.org/missing_sections) t
[17:38:21] <psinghal20>	 o provide better overview of how a campaign/course can target a particular category/topic. I couldn't seem to find any API documentation for this tool. Can anyone please point me to its API if it's available? Thanks!
[17:39:19] <lzia>	 psinghal20: hi. bmansurov will be able to point you to the right place. We are in a meeting now, so please expect some delay.
[17:39:27] <lzia>	 psinghal20: will be focusing on English first?
[17:41:33] <psinghal20>	 Thanks for the quick response! We are making the article finder compatible with different languages but as there are some which are available with only english wiki, we just hide those features for different languages.
[17:48:37] <leila>	 psinghal20: makes sense. We also do have some recommendations for French, and later hopefully more languages.
[17:49:26] <psinghal20>	 Yes, We are currently doing the same for ORES API for estimation scores and Page Assessment grades.
[17:59:21] <bmansurov>	 psinghal20: o/ Here's the code for the link you pasted: https://github.com/wikimedia/recommendation-tool and here's some documentation: https://www.mediawiki.org/wiki/GapFinder/Developers
[18:00:05] <bmansurov>	 psinghal20: I don't there's separate API documentation, you'll have to read the code unfortunately.
[18:00:23] <psinghal20>	 Hi @bmn
[18:07:35] <bmansurov>	 psinghal20: hi
[18:09:57] <psinghal20>	 Hi bmansurov, Thanks for the links. I tried looking through the developers guide and found one API mentioned there : https://github.com/schana/recommendation-missing-sections and I think the other link seems to be the code client side code written with react. Am I correct?
[18:10:58] <bmansurov>	 psinghal20: yes, you're right
[18:12:53] <psinghal20>	 Thanks! I will try skimming through the code to get a overview of the API. As I will be working with it, is there a way I can contribute to its API documentation?
[18:14:47] <bmansurov>	 psinghal20: I guess you can edit the Meta page, but mind you that that repo isn't being developed further.
[18:16:18] <psinghal20>	 Oh, is the research over this missing sections API still going on though? Also, is this API publicly accessible?
[18:21:54] <bmansurov>	 psinghal20: I think research is already done. Sure it's publicly accessible, but it's not production ready.
[18:24:59] <J-Mo>	 hi folks! The Research Showcase will begin in a few minutes. https://www.mediawiki.org/wiki/Wikimedia_Research/Showcase#June_2018
[18:27:58] <psinghal20>	 bmansurov: okay, I think it might still be usable for our project, I will take a look at the code. Is it okay if I ping you here if I have any other queries later on?
[18:28:25] <bmansurov>	 psinghal20: sure, feel free
[18:29:25] <leila>	 hello people. Please expect some delay for the start of the research showcase. The YouTube link is broken and brendan_campbell is setting up a new one.
[18:29:57] <brendan_campbell>	 https://www.youtube.com/watch?v=Q1sSzKKoHB8
[18:30:09] <J-Mo>	 ^^new YouTube stream link. Use this one instead.
[18:31:20] <J-Mo>	 HaeB can you tweet the new, improved YouTube link via @WikiResearch? https://www.youtube.com/watch?v=Q1sSzKKoHB8
[18:31:52] <J-Mo>	 we're almost ready folks; needed to make a last minute YouTube stream change. Please follow the link brendan_campbell just sent out
[18:32:01] <HaeB>	 J-Mo: what message do you suggest with it?
[18:33:22] <computermacgyver>	 halafak: Thanks for mailing the lists. You just beat me to it :-)
[18:33:59] <J-Mo>	 HaeB "this is the new link to the video. The previous link is no longer working, so use this one"
[18:34:01] <J-Mo>	 ?
[18:34:16] <computermacgyver>	 halfak: Opps can't type. Thanks for sending the link
[18:35:00] <HaeB>	 https://twitter.com/WikiResearch/status/1008780012600717312
[18:35:03] <J-Mo>	 and we are live!
[18:35:07] <J-Mo>	 thanks HaeB
[18:35:27] <HaeB>	 (can't join myself today unfortunately)
[18:35:44] <J-Mo>	 the stream should be starting soon, if it hasn't already. Feel free to ping me if you have problems, OR if you have questions for our speakers, Jonathan and Justine
[18:35:58] <computermacgyver>	 @J-Mo: it has started for me. Thanks!
[18:36:09] <J-Mo>	 awesome, computermacgyver
[18:37:17] <Giovanni_Ciampag>	 Hello!
[18:37:31] <J-Mo>	 hi Giovanni_Ciampag! Good to see you
[18:38:37] <Giovanni_Ciampag>	 J-Mo: good to see you folks!
[18:43:12] <miriam>	 ciao Giovanni_Ciampag!
[18:43:33] * Giovanni_Ciampag waves to Miriam
[18:47:29] <Giovanni_Ciampag>	 J-Mo: are you in charge with taking questions for the speaker?
[18:47:41] <J-Mo>	 Giovanni_Ciampag: yep
[18:48:33] <computermacgyver>	 Nice stat: out of ~50m conversations, only ~600 turn toxic with the existing participants
[18:48:48] <Giovanni_Ciampag>	 Here's my Q: how do you define in the first place what is a conversation?
[18:50:41] <J-Mo>	 got it, Giovanni_Ciampag
[18:51:12] <Giovanni_Ciampag>	 Tnx ;)
[18:51:15] <halfak>	 "So what you're saying is that you're a big dummy."
[18:51:36] <halfak>	 Basically, what paraphrasing like that always turns into. 
[18:51:37] <J-Mo>	 Giovanni_Ciampag: are you asking "how do you decide which of these talkpage threads are conversations, and which ones are not?"
[18:51:43] <Nettrom>	 Q for later: the filtering goes from 50M down to 635 pairs, is there a concern that this is too strict, e.g. that the first step from 50M to 3,000 removes potential good candidates?
[18:52:25] <ewhit_>	 How is toxic behavior operationalized here? Profanity, name calling, etc?
[18:52:26] <J-Mo>	 got it Nettrom
[18:52:27] <Giovanni_Ciampag>	 J-Mo: yes, as opposed to uncorrelated edits by multiple people on the same talk page
[18:52:27] <Nettrom>	 “too strict” is perhaps the wrong phrase there, maybe “there are issues with the initial prediction stage"
[18:52:44] <Nettrom>	 thanks J-Mo :)
[18:53:31] <J-Mo>	 got it ewhit_
[18:55:49] <Giovanni_Ciampag>	 J-Mo: another question: what about a conversation that has a personal attack in the middle but then recovers and goes back civil? Is that a conversation gone awry?
[18:56:17] <J-Mo>	 got it Giovanni_Ciampag
[18:59:57] <J-Mo>	 first presentation is wrapping up. any additional questions for Jonathan?
[19:00:00] <Nettrom>	 Q: is there a reason why the performance gain will come from getting the remaining 20% of false negatives right, or is there something to learn from the 28% that humans get wrong?
[19:00:40] <awight>	 ^ nice!
[19:00:42] <J-Mo>	 got it Nettrom
[19:00:46] <Nettrom>	 thanks
[19:00:48] <computermacgyver>	 Q: Could the speaker say a bit more about how they hope to move away from both the crowd and existing automated models to identify toxicity?
[19:01:00] <Giovanni_Ciampag>	 J-Mo: Oh, I guess that that answers my second question! :)
[19:01:25] <J-Mo>	 got it, Giovanni_Ciampag
[19:03:14] <leila>	 previous question was from Margeigh Novotny.
[19:03:27] <leila>	 Current question: Marshall Miller.
[19:08:14] <J-Mo>	 ewhit_, Nettrom: need to move on, but I will ask your questions (in that order) at the end of the second presentation
[19:08:39] <J-Mo>	 second presentation is beginning! Yiquing Hua is the speaker
[19:08:54] <J-Mo>	 *Yiqing, excuse me
[19:09:08] <Nettrom>	 no problem J-Mo, and thanks to the presenter for an excellent answer to my question! :)
[19:09:22] <ewhit_>	 Thanks J-Mo!
[19:09:43] <Giovanni_Ciampag>	 Thanks J-Mo!
[19:09:54] <computermacgyver>	 J-Mo: I had one in there too, but it can be dropped if there isn't time at the very end (or other more interesting questions come up!)
[19:09:55] <Nettrom>	 and yeah, thanks for keeping an eye on the chats, J-Mo :)
[19:10:47] <computermacgyver>	 12:00:48 computermacgyver | Q: Could the speaker say a bit more about how they hope to │ AntiComposite
[19:10:49] <computermacgyver>	                           | move away from both the crowd and existing automated       │ armalcolite
[19:10:51] <computermacgyver>	                           | models to identify toxicity? 
[19:11:25] <J-Mo>	 got it this time, computermacgyver
[19:11:52] <computermacgyver>	 J-Mo: Thanks. The other questions were more interesting in any case :-)
[19:19:24] <Nettrom>	 that breakdown overview is really nice!
[19:32:35] <Nettrom>	 hmmm, so what’s the probability that the conversation goes awry given its length? is there a variant of Godwin’s law at play here?
[19:33:08] <halfak>	 ^ love that question
[19:33:13] <halfak>	 Thanks Nettrom 
[19:33:17] <Nettrom>	 thanks everyone!
[19:33:21] <Giovanni_Ciampag>	 Thanks guys, really interesting talks!
[19:33:27] <Giovanni_Ciampag>	 Cheers!
[19:33:40] <Nettrom>	 great presentations and interesting work, looking forward to learning more about this in the future!
[19:33:42] <J-Mo>	 thank you all! see you next time :)
[19:34:10] <Ainali>	 Nettrom: Isn't that exactly Godwin's law?
[19:34:38] <Nettrom>	 Ainali: no, Godwin’s law is that the probability of a reference to the nazis goes to 1 as thread length increases
[19:34:52] <Nettrom>	 I was generalizing it to not specifically refer to nazis
[19:35:02] <Ainali>	 Well, I would argue that is a conversation gone awry
[19:35:06] <Nettrom>	 of course :)
[19:35:57] <Ainali>	 Ah, so you would say while some conversations mentions nais, but some go awry earlier? :)
[19:36:59] <Nettrom>	 that seems also to be true, yes
[19:37:47] <Ainali>	 We might have a special Godwin's law and a general Godwin's law :)
[19:38:32] <Nettrom>	 I had to look this up again, Godwin’s specifically references Hitler
[19:39:45] <Ainali>	 General: As an online discussion grows longer, the probability of going awry approaches 1.
[19:40:00] <awight>	 There's a nice postscript too, when Godwin pointed out that it's actually non-absurd for a conversation to reference Nazis when you're talking about people acting like Nazis.
[19:41:54] <Nettrom>	 the Wikipedia page links to an archived EFF page with a bunch of variations and such: https://web.archive.org/web/20150627165048/http://w2.eff.org/Net_culture/Folklore/Humor/godwins.law
[19:44:32] <Nettrom>	 oh well, I should be writing a paper review, not researching Godwin’s law, haha…
[23:12:33] <halfak__>	 moar underscores!