[00:01:14] articlequality seems to be in sync, still pulling down all the lfs objs for editquality [00:14:41] pushed up 2 new objects for editquality on gerrit [00:22:56] 10ORES, 10Scoring-platform-team (Current), 10Gerrit: Write a cookbook for the workaround for getting LFS to gerrit - https://phabricator.wikimedia.org/T226055 (10ACraze) LGTM I followed the LFS commands and pushed 2 new objects to gerrit for editquality. It seems like articlequality and draftquality are bot... [14:36:10] o/ kevinbazira [14:36:14] How's hacking? [14:44:29] 10Scoring-platform-team, 10Research: Extract cross-wiki WikiProject tags - https://phabricator.wikimedia.org/T240273 (10Halfak) https://figshare.com/authors/Aaron_Halfaker/96516 [14:55:34] o/ halfak [14:56:26] It's going well, I generated the skipgram 100 cell vector model and it's word vectors [14:56:46] 10Scoring-platform-team, 10Discovery-Search, 10Growth-Team: Allow searching articles by ORES drafttopic - https://phabricator.wikimedia.org/T240517 (10Halfak) I don't think that ORES is the right place to do any joining of data. ORES output format doesn't allow us to append a QID to it. The revision-scor... [14:56:47] Nice! [14:56:56] Do you have the 50 cell ones too? [15:00:24] Yes I do [15:03:50] the 50 cell ones completed but the 100 cell ones are still running [15:05:10] Gotcha. Great! Where can I find the files? [15:05:17] I'd like to dig into them a bit. [15:05:51] https://phabricator.wikimedia.org/T235184 [15:06:01] ^ For posting details. [15:15:16] kevinbazira, ^ [15:30:53] I have added details of where to find the files on the phabricator link above [15:31:10] Thank you! [15:34:24] You're welcome :) [15:46:48] I wonder if there are some good ways we could trim this .vec file -- e.g., by removing either super rare or super common words. [16:10:33] * kevinbazira is going AWK, will be back in time for async [16:31:50] (03CR) 10jerkins-bot: [V: 04-1] build: Updating mediawiki/minus-x to 0.3.2 [extensions/ORES] - 10https://gerrit.wikimedia.org/r/558584 (owner: 10Libraryupgrader) [17:02:36] async standup: [17:02:47] Y: Continued work on Jade UI components & widgets. Got facets and proposals rendering correctly (minus css styling). Also reviewed the git lfs docs for manually syncing our models on gerrit. [17:03:12] T: More Jade UI, specifically endorsement & author widgets, proposal edit menu and maybe 'propose new label' popup + form [17:03:38] Y: generated skipgram 50 cell vector model and did some ML tutorials [17:03:52] T: started process of generating skipgram 100 cell vector model and shared details with halfak in a comment here: https://phabricator.wikimedia.org/T235184 [17:06:50] Y: Rebuilt the wikilabels instances that were still running Debian Jessie (found a scary backup bug!) Re-trained the enwiki and dewiki models. Started work on a beta deployment of ORES. Submitted a patch for fixing Jade API's i18n QQQ. Got my figshare.com account back so I could upload drafttopic datasets. [17:06:50] T: Mostly meetings today. Will try to get the draft topic streams merged and possibly even a new drafttopic model built. [17:07:14] I'm blocked on review for https://github.com/wikimedia/editquality/pull/218 [17:08:40] i can review that in a bit halfak [17:08:45] Thanks! [17:09:01] With that, I should be able to get an ORES deployment out to beta today between meetings. [17:09:36] kevinbazira, next step for vectors is to implement a strategy for loading them in and generating features. [17:11:01] See https://github.com/wikimedia/revscoring/blob/master/revscoring/datasources/meta/vectorizers.py [17:11:17] We have one for word2vec and we should probably have one for fasttext. [17:11:28] Maybe it would make sense to read the vectors in ourselves. [17:11:36] Great. I'm going to work on that [17:11:38] It's a pretty simple file format. [17:11:41] Cool :) [17:18:51] me is going AWK, good day halfak and accraze [17:19:02] Have a good evening! [17:22:39] see ya later kevinbazira [17:33:44] halfak, i just reviewed and merged those revscoring 2.6.2 rebuild PRs for editquality and articlequality [17:33:52] Thank you :) [17:33:54] anything else you need for the deploy? [17:34:13] Not right now. Might ping if I run into something. Hopefully I'll just have you reviewing the config changes. [18:01:32] hey groceryheist [18:01:36] Are we meeting now? [18:01:48] I don't have a call to join. [18:02:36] oh hey [18:03:05] let me get a link [18:03:47] https://beta.meet.jit.si/mako_aaron_nate [18:03:50] mako's on his way [18:05:46] halfak: ^ [19:00:29] groceryheist, sorry had to run to next meeting [19:46:57] halfak: do you need/want me there for "Using the ORES topic model in UX" meeting this afternoon? if not, i will likely skip as I doubt I would have anything unique to provide [19:50:14] Na. I think I can handle it. Will report back [19:50:17] isaacj, ^ [19:50:31] :thumbs up: thanks! [21:22:18] (03CR) 10Umherirrender: [C: 03+2] "Resubmit" [extensions/ORES] - 10https://gerrit.wikimedia.org/r/558584 (owner: 10Libraryupgrader) [21:24:29] hey halfak is there a way to get a copy of the 30 day ores score database? [21:25:16] I have some missing data based on the historical scores and I want to cross-reference with the live site [21:27:14] halfak: also can you share artifacts where Wikipedians talked about using anon and redlinks in patrolling work? [21:27:55] groceryheist, hey! I can get that to you late this week. [21:28:08] Re. the last 30 days of scores, you can get that out of some tables accessible to quarry. [21:29:22] https://quarry.wmflabs.org/query/40711 [21:29:27] groceryheist, ^ [21:29:58] Nettrom, any chance you remember some conversations with Wikipedians about suggestbot turning redlinks to bluelinks for newcomers? [21:30:08] I remember that came up for you. I'll just dig through your talk page. [21:30:16] ... if you don't have time :) [21:30:38] * halfak looks around for Vermont. [21:32:42] halfak: which wikis are those revisions in quarry from? [21:32:59] Oh those are from enwiki. [21:33:07] Enwiki is the default if you don't specify [21:33:26] Otherwise, you can do "USE arwiki_p;" before the query to select a different wiki. [21:37:43] ah ok [21:38:48] and what if I want a random sample say? [21:39:04] never mind [21:39:07] I can figure that much out [21:43:22] thanks [21:50:00] o/ accraze [21:50:08] found another PR I need: https://github.com/wikimedia/ores/pull/334 [21:50:14] Should be easy. [21:58:48] halfak it's merged! [22:15:35] halfak: I don't remember off the top of my head whether those discussions happened on SuggestBot's or my talk page [22:16:10] but I'm pretty sure that redlink recommendations has been discussed at some point [22:21:57] Aha. This is about redlink user pages. Not redlink articles. [22:45:40] As in if suggestbot posts recommendations to a user's talk page, it'll turn blue. [22:45:52] I think you were looking at running a newcomer experiment with suggestbot or something like that. [22:46:32] git review is taking a LOOONG time on stat1007 [22:46:35] * halfak yawns [22:54:13] (03PS1) 10Halfak: Updates for revscoring-2.6.2 [research/ores/wheels] - 10https://gerrit.wikimedia.org/r/558715 [22:54:47] (03PS2) 10Halfak: Updates for revscoring-2.6.2 [research/ores/wheels] - 10https://gerrit.wikimedia.org/r/558715 (https://phabricator.wikimedia.org/T240725) [22:54:56] accraze, https://gerrit.wikimedia.org/r/#/c/research/ores/wheels/+/558715 [22:56:41] halfak: hm, maybe that was discussed in the bot request for the experiment, let me check [22:57:02] Thank you :) [22:57:04] I know I also looked into whether users in Norwegian Wikipedia were welcomed if SuggestBot had posted to their talk page as part of the experiment, and the answer was always "no" [22:57:15] interesting! [22:57:24] Is that published? [22:57:40] no, that was the newcomer recommendations project that we never wrote a paper for :( [22:57:55] (03CR) 10Accraze: [C: 03+2] Updates for revscoring-2.6.2 [research/ores/wheels] - 10https://gerrit.wikimedia.org/r/558715 (https://phabricator.wikimedia.org/T240725) (owner: 10Halfak) [22:57:56] gotcha. [22:57:58] \o/ [22:58:01] <3 accraze [23:02:28] (03PS1) 10Halfak: Updates for revscoring 2.6.2 and some models. [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/558719 (https://phabricator.wikimedia.org/T240725) [23:02:44] accraze, https://gerrit.wikimedia.org/r/#/c/mediawiki/services/ores/deploy/+/558719 [23:03:03] this one is a bit more complicated. If you find time for it today, I'll get a beta deployment out tomorrow. [23:04:14] halfak: I don't find discussions of turning new user links blue in either the bot request or the discussion on the Welcoming Committee's pages (https://en.wikipedia.org/wiki/Wikipedia:Bots/Requests_for_approval/SuggestBot_3 and https://en.wikipedia.org/wiki/Wikipedia_talk:Welcoming_committee/Archive_4#Experimental_research_study_targeting_newly_registered_users ) [23:04:29] Gotcha. Thanks for looking. [23:04:36] I'll go do some more digging elsewhere. [23:04:42] there were questions about how it would handle users who were welcomed with a warning, or if it would welcome users as well [23:07:06] cool, taking a look at it now halfak [23:19:38] Nettrom: do you have some documentation of this? [23:20:53] (03CR) 10Accraze: [C: 03+2] Updates for revscoring 2.6.2 and some models. [services/ores/deploy] - 10https://gerrit.wikimedia.org/r/558719 (https://phabricator.wikimedia.org/T240725) (owner: 10Halfak) [23:21:37] halfak: i forget, is there no jenkinsbot to do verify on the deploy repo? [23:23:17] groceryheist: documentation of the research project? [23:24:55] yeah and specifically of people's concern about red links [23:29:10] Netrom oh i'm seeing the links you posted above [23:29:30] I have a googledoc with a summary of the experiment somehwere too, let me see what's in it [23:32:09] ok yeah it doesn't seem like people were explicitly concerned about turning redlinks blue. [23:32:57] maybe we'll need to run a quick survey if we need to find evidence that people use anon or redlinks to find edits to review.