[17:54:11] Hello everyone. [17:54:15] I'll be the IRC factotum again today. [17:54:17] Metrics will begin in a few minutes. [17:57:11] James_F, I see the stream link isn't in the /topic. Is that on the way? [17:57:42] halfak: Done. [17:57:46] Thanks [17:58:30] Now with one fewer Sheffer. [18:00:45] About to start. [18:00:50] o/ [18:00:55] Hey kids [18:01:18] There is goes. [18:01:19] (As always, if anyone has any questions please shout them out; we've got Q&A at the end.) [18:01:27] Awesome background photo. [18:02:02] Welcome, everybody. :) [18:02:22] Somehow I always get missed on the anniversaries. [18:02:37] halfak, are you now? [18:02:37] I'm at 4 years if you count my contracting work [18:02:58] September should be the right month. [18:03:12] They did for Ocaasi, so happy anniversary. [18:03:20] What did Luis say about "lulls"? [18:03:27] what is F&A? Not on https://wikimediafoundation.org/wiki/Staff_and_contractors [18:03:41] Fundraising & Advancement? [18:03:52] Finance & Admin, probably. [18:04:16] Yes, Finance and Administration. [18:04:42] Staff & Contractors says " [18:04:46] "Finance, Administration, and Human Resources" [18:05:11] and IT! [18:05:17] Outreacy and GSoC [18:05:23] Outreachy [18:05:26] :) Awesome to hear people are liking VE. Congrats to the VE team, past and present. [18:05:27] "human resources" always makes me think they're planning to turn us into soylent green :D [18:05:28] spagewmf: Are you still surprised by the lack of consistency of our internal structures? [18:05:29] F&A was F&A for a bunch longer than it was a sister of HR and OIT, so they couple them together. [18:05:49] brion: Tsk, that's FY2018 outline planning only! ;-) [18:05:55] lol [18:05:59] spagewmf: You must be new here :) [18:06:02] guillom: yes, also I can't parse a TLA that isn't a three-letter acronym :) [18:06:09] yay, code of conduct [18:06:13] spagewmf: "ETLA". :-) [18:06:22] +1 [18:07:49] (Questions for what Luis was talking about now we're moving on to Dan?) [18:08:19] Discovery has a nice-looking dashboard... [18:09:29] And a lot of cool work going on. [18:09:40] Indeed. [18:11:00] brion: hence the HR logo (which I can't find): they fertilize us, fatten us up, then... "WMF is people!" 8-D [18:11:26] i always thought the free food after metrics meetings was suspicious... [18:12:03] Maybe we should add a 'time spent fetching your results' to the search results, like Google. That should make it seem faster... [18:12:05] ;) [18:12:36] http://searchdata.wmflabs.org/metrics/ [18:12:59] I have a strategy question about search, when there's time for such a question [18:13:01] Also https://www.mediawiki.org/wiki/Wikimedia_Engineering/2015-16_Q1_Goals#VisualEditor_Team was the URL in the corner of Dan's slides. [18:13:19] woo hoo -- typo fixing! [18:13:53] James_F: are you manning the mic again? [18:14:10] Pine: No, I'm IRC factotum, other people are manning the mics. [18:14:15] Also, personing? [18:14:20] hmm, I wonder if the action=search API does this substituted search [18:14:20] * Pine waves to try to flag down a mic [18:14:36] Pine: You're on IRC… [18:14:43] pine: james is your conduit for questions via irc [18:14:45] Thanks James_F for the Visual Editor link...was just looking for it! [18:15:00] Pineapples100: Thank you. It's very much a team effort. :-) [18:15:23] what's a DOI ? [18:15:34] Digital Object Identifier. [18:15:43] James: let me rephrase my question. How can convey my question so that it can be converted into audio format to Dan? [18:15:44] It's like an ISBN for digital papers. [18:15:50] nice [18:15:53] like a URL for digital papers [18:15:59] Pine: Ask it here. :-) [18:16:30] DOI doesn't necessarily mean you can read it online for free, for clarity. [18:16:43] completion suggester! [18:17:06] disable search! [18:17:12] oh :) [18:17:15] Question: can someone describe how the strategy update affects the work happening in Discovery? [18:17:29] omg finally autocomplete! [18:17:35] yay [18:17:41] can haz in apps now please? [18:17:56] Pine: Which strategy update? The one from Lila a few months ago? [18:18:12] bgerstle: I'll ask that too. [18:18:23] James_F: no, the update that was just noted in the blog, from Terence. [18:18:31] Those redirects are not *just* for Wikipedia built-in prefix search. [18:18:34] Could you identify the strategy update you are referring to by DOI, please? :) [18:18:44] James_F: i guess the question would "what's the timeline for when autocomplete will be in production" [18:18:44] They also affect direct URL navigation (which I use all the time), and maybe Google. [18:18:51] ooooh good autocomplete to replace the old prefix autocomplete? awesome sauce :D [18:19:13] I'm confused by http://suggesty.wmflabs.org/wiki/Jurrasic_park [18:19:17] Pine: Oh, you mean the strategy consultation results that were presented at Metrics two months ago? http://blog.wikimedia.org/2015/08/27/strategy-potential-mobile-multimedia-translation/ [18:19:25] jaufrecht: I'm afraid that WMF's current search engine doesn't let me search for the DOI for a blog entry [18:19:33] bgerstle: Yeah. [18:19:48] Pine, sure it does. You'll just get zero results. ;) [18:20:17] James_F: yes, that blog entry [18:20:32] Pine: Cool. [18:20:41] matt_flaschen: you mean an https://en.wikipedia.org/wiki/%s bookmarklet ? Me too. [18:21:15] spagewmf, basically, but I usually just change the last part instead of using a bookmarklet. [18:21:39] bgerstle, Pine: Your questions are in the queue, worry not. [18:22:05] np, not very pressing. seems like they're working towards getting it out [18:22:09] depending on results of A/B testing [18:22:13] ok [18:22:15] * James_F nods. [18:22:18] We need to show this map to anyone who thinks Wikipedia is mostly done. [18:22:24] matt_flaschen: Indeed! [18:22:42] matt_flaschen: I like the Wikidata comparative charts for that too. [18:22:53] limited content in your language -> Reasonator and AutoDesc! To quote Magnus Manske, it's a 250x force multiplier [18:23:08] :) maybe we should change the logo to have more missing puzzle pieces [18:23:09] *misquote :) [18:23:18] to compensate, en.wiki has good coverage of all the Star Trek planets and civilizations... [18:23:37] milimetric: Or it should animate to add a piece every time you write a new article? [18:23:39] qgil: We're gonna need a bigger map :) [18:24:03] qgil: so we're done then? we can all go home? [18:24:04] James_F: I'm on it, /me pulls out his macromedia flash editor [18:24:07] Hmm, yeah, how many of those geo-located items are for fictional places? [18:24:12] milimetric: Gah. :-0 [18:24:21] :-P [18:24:28] E.g. Star Fleet Academy is in San Francisco. [18:24:32] Does that mean it's geo-tagged? [18:24:33] bgerstle, I hope the "..." in my sentence expressed my sadness about thisa [18:25:13] qgil: yeah [18:25:22] James_F, Q for Leila, how did you make estimates about the amount of time it would take the get that coverage? [18:25:27] interesting maps. [18:25:29] halfak: Thanks. [18:26:38] neat! -- nudge users to create new pages in other languages on interests they have.. [18:26:56] the missing content IS THERE in your language on Wikidata! productize Reasonator and AutoDesc FFS. [18:27:02] what are bebal templates? [18:27:10] subbu|lunch: Babel [18:27:10] James_F: Question for Leila: How does it determine important content? [18:27:18] subbu|lunch, Babel. Templates that say which languages you know. [18:27:18] * James_F nods. [18:27:25] ah, babel, ok. [18:27:29] https://meta.wikimedia.org/wiki/User_language [18:27:29] James_F, Follow-up Q if there's time. By directing editors' attention to increase coverage, we are removing their attention from other activities. How do we know that this is a better use of Wikipedian's time? [18:27:29] User:E is a real user :P [18:27:43] Got them. [18:27:55] * guillom smells food. [18:28:13] matt_flaschen: Babel's a parser function now, right? [18:28:19] Not just a template. [18:28:22] Yeah [18:28:27] all this sounds like a good approach [18:29:05] Not sure if all the Babel templates actually use #babel though. [18:29:14] Probably not. :-) [18:29:33] halfak, but, editors decide though, isn't it? [18:29:48] subbu|lunch, sure, but this is an issue with task routing in general. [18:30:19] If we direct you somewhere saying that your time/energy would be better spent *there*, we're implicitly saying that it's less useful where you are currently putting it. [18:30:30] I'm not sure we have a good theoretical framework for coming to that conclusion. [18:30:36] I'd like to hear Leila's thoughts. [18:30:42] It's something I worry about getting right when considering adding suggestions to VisualEditor. [18:31:00] "Hey, before you save you should add a reference" probably does encourage good moves. [18:31:00] ok. i'll keep my thoughts to myself for now. curious what leila has to say [18:31:02] "Probably". [18:31:21] subbu|lunch, :( sorry if I came off bad. [18:31:26] no, no. :) [18:31:31] But it also means people don't save and move on to the next page to make the edit they were intending, necessarily. [18:31:41] Maybe this is OK. Maybe it's fantastic. Maybe it's a loss. [18:31:46] Knowing which is hard, but important. [18:32:07] James_F, +1 good thoughts. It would be very interesting to experiment with :) [18:32:14] James_F: how do we know that article in language 1 does not exist language 2.. (imperfect translation, etc.) [18:32:18] halfak, definitely some of the recommendations will remove attention from other WP activities. But some may just cause an overall increase in a user's Wikipedia activity. [18:32:24] halfak: Indeed. :-) [18:32:36] Q: how do they correlate article titles between languages? Presumably we aren’t asking French editors to write an article titled ‘Earthquake Warning System' [18:32:37] cajoel: I think it's based on Wikidata cross-language links. [18:32:41] matt_flaschen, fair point. [18:32:44] andrewbogott, Wikidata [18:32:53] matt_flaschen: Ah, ok [18:33:00] matt_flaschen: With pre-written titles? Nice. What about clashes? [18:33:12] James_F, ask Leila :D [18:33:15] matt_flaschen: E.g. "John Smith" is the Wikidata label for a dozen people… [18:33:19] * James_F will do. :-) [18:33:23] James_F, I don't know about pre-written titles. You *could* do that in some cases with labels, as you note. [18:33:24] Question: Will we be suggesting 5 possible articles to translate/edit for users? why not 2 or three? [18:33:31] Kk. [18:33:50] I wonder if ContentTranslation helps with disambig titles. [18:34:25] matt_flaschen, andrewbogott, victorgrigas: Got your questions. [18:34:56] matt_flaschen: yes, (I'm working on the prototype for their tool), they use ContentTranslation for title mapping [18:34:59] James_F: what is the protocol for questions from people who are local? [18:35:11] ori: Wait 'til the end when questions are being asked. [18:35:15] nod [18:35:32] milimetric, do you know how CT does title mapping? [18:35:38] ori: Do you mean, will I ask your question for you? Sure, if you'd prefer. [18:35:40] I do not, we should ask them [18:35:44] James_F: nonono [18:35:55] ori: OK [18:35:58] :-) [18:35:58] does anyone know what provides "viewed 2545 times recently" on an article? I thought the page view firehose was too overwhelming to count [18:36:02] I must vocalize it with my vulgar american accent [18:36:11] * James_F grins at ori. [18:36:17] LOL [18:36:26] ori: Question-launderer since 2005. ;-) [18:36:26] spagewmf: where is that shown? [18:36:31] spagewmf: ellery crunches that in hadoop and the data is quite out of date, but he crunches it when he makes the models [18:36:55] Now that we have voting browser tests, if release engineering spends all their time going forward building huge trebuchets, I'm surprisingly okay with that. [18:37:19] :-) [18:37:26] I think spagewmf is talk ing about stats.grok.se, for example http://stats.grok.se/en/latest/Troubled_Asset_Relief_Program [18:37:31] spagewmf: but the pageview API is coming along nicely, so we'll use that when it's up in the next month or so [18:37:40] ebernhardson: Leila's slide showed a half-dozen wiki pages, each with that phrase at the bottom [18:37:40] this "Feedback" graph is beauthiful [18:37:50] milimetric: is there anywhere i can access this out of date information? we really want to try integrating page view information into the completion scoring(thing dan showed earlier in meeting). The completion index isn't even updated in real time (expecting once a week batch process) so out of date is probably still plenty relevant [18:37:52] Pine: naw, ellery doesn't go to stats.grok.se for this [18:38:35] ebernhardson: right now, just in hadoop, there's a small-ish Hive table called "wmf.pageview_hourly" [18:38:47] ebernhardson: but soon, in the API [18:38:51] Soon. [18:39:20] really though, I know other people have promised this and not delivered. I'm promising it this time :) [18:39:29] milimetric: that hourly table doesn't sound like "viewed 2545 times recently" [18:39:39] its certainly on me and not you, but since yours is done at the end of the quarter thats too late to integrate the answer into my quarterly goals :P [18:39:43] milimetric: Release the Kraken ;) [18:39:46] spagewmf: you can query back for as much history as we have, by article title, and sum [18:39:57] By the way, James_F : thanks for the features for automatically linking ISBNs and PMIDs. [18:40:04] spagewmf: the Kraken has sunken back to the depths from which it came. Our approach is much cuddlier [18:40:15] Pine: Happy you like it. Next week it will have a proper editor for them which makes it a bit cleaner. [18:40:22] milimetric: I was telling ebernhardson that we were planning to stand a test instance of the API on labs [18:40:24] Pine: Further suggestions welcome. [18:40:25] Cool. Are DOIs in the queue? [18:40:28] he answered the qn. i was gong to ask. [18:40:45] Pine: No, we're probably not going to add "DOI 1000.12345.x" auto-linking to MediaWiki. [18:40:56] Any reason why not? [18:41:03] Pine: But we're stuck with ISBN/RFC/PMID due to brion deciding it too long ago. [18:41:06] these slides and images are fantastic [18:41:10] yaaaay [18:41:22] auto-linking should be done by the editor, not the markup parser [18:41:28] now that, you know, we have an editor [18:41:53] madhuvishy: cool, but we're probably not going to keep that updated enough for prod use. ebernhardson: we should talk and see if it's possible for Hadoop to back up your use case. Set up a meeting [18:42:00] Pine: Changing wikitext syntax is always a breaking change. Potentially it could damage things. There's not enough pay-off to make it worthwhile. [18:42:16] How are you measuring payoff? [18:42:16] milimetric: Yup, just as a preview [18:42:38] Pine: Professional judgement on the value of spending volunteer time scanning millions of pages for breakages. [18:42:41] James_F: Like brion said, it's not required to change wikitext syntax. If auto-linking DOI as you type it is useful in certain contexts, the editor can do that without changing markup. [18:42:59] hm... we can discuss more later offline if I remember. :) [18:43:04] matt_flaschen: Yup, but that's not what Pine is asking. [18:43:15] Pine: Sure. :-) [18:43:46] Pine, are you talking about in VE or in the wikitext editor? [18:43:55] matt_flaschen: VE [18:44:33] Oh, in that case, possibly. [18:44:42] But without a DOI search index it'd be rather low value. [18:44:56] * Pine gives a long stare at Discovery [18:45:08] * Pine wonders when I can search for Wikimedia blog entry DOIs [18:45:14] * harej gives a long stare at Pine [18:45:19] Well, cite doi is quite useful (though maybe not as necessary now we have Citoid). Not sure how useful auto-linking DOIs are otherwise (why not just use cite doi)? [18:45:40] * legoktm claps [18:45:41] Don't think Wikimedia blog has a DOI. Maybe I'm wrong, though. :) [18:45:55] * matt_flaschen applauds voting end-to-end/browser tests [18:46:04] Also, where would it link? [18:46:12] I read evidence somewhere on the internet (maybe Steve McConnell?) that no form of testing, manual or automatic, catches more than about 50% of bugs, but code review was good for 60%. with, presumably, limited overlap. [18:46:26] James_F: dx.doi.org/ ? [18:46:27] James_F, in practice, everyone uses http://dx.doi.org/ [18:46:36] Possibly. [18:47:03] I wonder what the median amount of code review is across teams. it's 100% for VE, right? [18:47:13] 100% for analytics [18:47:19] It's meant to be 100% for all devs. [18:47:22] Not for Ops though. [18:47:27] For hysterical raisins. [18:47:39] i hope not throwing it away though :) [18:47:46] it seems promising right now, at least [18:47:53] well, measuring CR is hard -- do you mean 'a second person +2d the change' or 'someone carefully checked every line of coode for potential issues' :D [18:48:10] I can't recall hearing the phrase "hysterical raisins" in Office before. There's something new every day around here. [18:48:20] brion, or 'carefully checked every line and carefully manually tested edge cases' [18:48:23] brion: we mean "someone checked every line of code". But I'm not sure about "carefully" :) [18:48:25] I would imagine that only the latter would catch bugs at a rate comparable to testing [18:48:57] Wait... so # of articles coverts to years directly? [18:49:14] I've been wasting my time on labor hour measures! [18:49:35] halfak: As an approximation across the whole population, maybe it's good enough? I dunno. [18:49:44] halfak: be sure that you value editors' time at $300 per hour in your editor productivity measures. [18:49:59] Not IMO. I think a method should be stated or they should say they made it up. [18:50:15] Pine, na. Lowly grad students. $20 per hour for 1/6th the hours you actually work. [18:50:23] It could still be diverting from other tasks, such as copy-editing, fixing unverifiable content, etc. [18:50:28] ^ [18:50:30] That [18:50:54] But I firmly believe good recommendations will also divert people from spending time on Facebook. [18:50:59] And it's not clear which is a greater effect. [18:50:59] halfak: I think Pine is referring to recent events about paid editing and their nominal rate. [18:51:06] Oh! [18:51:17] At $300 per hour, I need to switch jobs [18:51:40] $300/hr is retail/extortion price, not wholesale [18:51:42] I'm not sure US$300 an hour is good enough for felony work. [18:51:57] studies show crack dealing pays below minimum wage [18:51:58] My charge rate for criminal activities are in the billions an event. [18:52:18] (So I can redistribute to make up for my activities, obviously.) [18:52:20] Wikidata is working on that I think [18:52:26] I'm pretty sure Reasonator does not support every language in the world. [18:52:29] It's someone's batchelor thesis? [18:52:35] Gotta hire lawyers [18:52:51] If there were, a lot of companies would be offering Magnus a lot of money for that technology. [18:52:58] But Reasonator is cool. [18:53:03] https://www.wikidata.org/wiki/Wikidata:Article_placeholder_input / https://meta.wikimedia.org/wiki/Filling_red_links_with_Wikidata [18:54:32] Pine: I need to have a chat with my boss if that's the going rate for Wikipedians now [18:55:00] hi harej :) [18:55:13] Hello dynosaur [18:55:52] is @spagewmf in the house? [18:56:08] matt_flaschen: that's what happens when one person writes it, and it uses tech outside WMF, and automatically generating an aricle falls neither in Discovery, Reading, or Editing [18:56:27] That's us done. [18:56:29] spagewmf: what's the name of the tool you were referring to? [18:56:30] Thank you all. [18:56:47] SVentura: https://tools.wmflabs.org/autodesc [18:56:48] spagewmf, my point was more that Reasonator is not replacement for articles, for many reasons (it not supporting every human language is just one of them). [18:56:55] spagewmf: thanks! [18:56:56] wait, didn't ori have a question? [18:57:04] it was answered [18:57:07] milimetric: He did but then it was… yeah. [18:57:11] oh, oops, I missed it somehow [18:57:18] spagewmf: who leads this? [18:57:23] milimetric: Answered without him asking. [18:57:27] ah [18:57:35] matt_flaschen: of course, but c'mon!!!! There are wikis with a few thousand articles. They could go to millions with some effort on AutoDesc and Reasonator [18:58:28] spagewmf, that would not necessarily be healthy for the communities future development. It's a complicated question. Showing a Reasonator-generated blurb when there is no article (distinct from actually auto-editing the DB) is a different idea. [18:58:54] James_F: if you need an impact statement for your boss or whatever, the government agency I work for uses VisualEditor in its training materials exclusively. I consider wiki text to be deprecated and not worth teaching for my purposes. [19:00:14] matt_flaschen: Oh, never create an article in the DB, always create it and cache it in RESTBase. The bar should be "Write a better article than this automatic thing that continually gets better". That's a win outside the large-wiki-centric fantasy of thousands of editors polishing tens of thousands of articles [19:00:32] s/always create/always generate on-the-fly/ [19:01:06] Compromise: put resonator-like data in the editing window [19:01:16] spagewmf, yeah, that version is worth talking about. [19:01:19] Rather than on the "reading" side [19:02:30] harej: so someone looking for information articles in Inuit doesn't see anything? Doesn't sound like free open knowledge to me.\ [19:03:10] harej: also, automatically-generated descriptions and stubs get better over time thanks to improvements in Wikidata and to the generator code. [19:03:13] spagewmf: the compromise is based on the idea that auto generation of articles hurts article creation [19:05:52] harej: whatever the [[template]] is for "more study needed" :) If a robot description clearly flagged as such with a pencil next to it discourages people from writing something better, that seems completely outweighed by having 1 million articles in your native language instead of 50,000 [19:08:43] spagewmf: have you seen https://www.wikidata.org/wiki/Wikidata:Article_placeholder_input ? [19:18:54] James_F: citoid saves taxpayer money. I'm not even kidding [19:27:21] harej: :-) [20:11:57] legoktm: thanks, I see https://www.wikidata.org/wiki/Wikidata:Article_placeholder_input#We_already_have_most_of_what_it_takes_in_ReasonatorReasonator_DOES_provide_information_in_any_language_about_ANY_subject_that_has_a_Wikidata_item. So which vertical owns article placeholder? I would argue Discovery. I suspect on an wiki with few articles, people try search, see little, and give up. Those w [20:12:03] ho speak a more popular language as well just visit it instead. [20:16:14] No WMF vertical does... [20:19:08] Any vertical could own it if they wanted to [20:58:02] feels like reading or discovery [20:58:05] at leat in theory :D [21:02:25] spagewmf: I believe the Wikidata team is working on something about this. They are in touch with us in Discovery about numerous things, including this. :-) [21:04:48] Deskana: yay! As you may know, Reading has some interest in AutoDesc, since the Wikidata description appears in search results and over lead images. Dimitry Brandt has a fork of AutoDesc [21:05:50] Deskana: but it's a different use case -- better "signposting" of an existing article -- rather than providing information when there's no article