[00:14:30] halfak: you brought up the "undo" messages that are left when people click "revert" in recent changes. Is there a way you can think of to find out what these messages would be? [00:47:50] ok I think I solved one of the problems I was having with spark [01:49:02] 10ORES, 10Scoring-platform-team, 10Analytics, 10Dumps-Generation, and 3 others: Decide whether we will include raw features - https://phabricator.wikimedia.org/T211069 (10Milimetric) +1, either hops is doing some incredible organic marketing or their ideas and libraries are good and people are using it. I... [06:16:43] 10Scoring-platform-team, 10Diffusion, 10editquality-modeling, 10Release-Engineering-Team (Backlog), 10artificial-intelligence: Gerrit repo scoring/ores/editquality not mirroing - https://phabricator.wikimedia.org/T224996 (10hashar) [06:17:19] 10ORES, 10Scoring-platform-team, 10Analytics, 10Dumps-Generation, and 3 others: Decide whether we will include raw features - https://phabricator.wikimedia.org/T211069 (10awight) >>! In T211069#5229195, @ArielGlenn wrote: >>>! In T211069#5229037, @awight wrote: >> Another production feature store framework... [07:05:49] PROBLEM - check load on ORES-web01.Experimental is CRITICAL: connect to address 172.16.3.131 port 5666: Connection refusedconnect to host ores-web-01.ores.eqiad.wmflabs port 5666: Connection refused [07:06:03] PROBLEM - check users on ORES-web01.Experimental is CRITICAL: connect to address 172.16.3.131 port 5666: Connection refusedconnect to host ores-web-01.ores.eqiad.wmflabs port 5666: Connection refused [07:07:37] PROBLEM - check disk on ORES-web01.Experimental is CRITICAL: connect to address 172.16.3.131 port 5666: Connection refusedconnect to host ores-web-01.ores.eqiad.wmflabs port 5666: Connection refused [07:10:21] PROBLEM - puppet on ORES-web01.Experimental is CRITICAL: connect to address 172.16.3.131 port 5666: Connection refusedconnect to host ores-web-01.ores.eqiad.wmflabs port 5666: Connection refused [07:23:37] RECOVERY - check disk on ORES-web01.Experimental is OK: DISK OK [07:23:49] RECOVERY - check load on ORES-web01.Experimental is OK: OK - load average: 0.11, 0.28, 0.26 [07:24:03] RECOVERY - check users on ORES-web01.Experimental is OK: USERS OK - 1 users currently logged in [07:24:55] RECOVERY - puppet on ORES-web01.Experimental is OK: OK: Puppet is currently enabled, last run 1 minute ago with 0 failures [07:33:05] 10ORES, 10Scoring-platform-team, 10Analytics, 10Dumps-Generation, and 3 others: Decide whether we will include raw features - https://phabricator.wikimedia.org/T211069 (10Nuria) >Please do let me know what the current consensus is around postponing any feature pipeline work until WMF has a dedicated machin... [13:43:41] o/ [14:03:08] Technical Advice IRC meeting starting in 60 minutes in channel #wikimedia-tech, hosts: @CFisch_WMDE & @bd808 - all questions welcome, more infos: https://www.mediawiki.org/wiki/Technical_Advice_IRC_Meeting [14:46:06] 10Scoring-platform-team, 10articlequality-modeling, 10artificial-intelligence: Create new labeling campaign for Basque Wikipedia articlequality - https://phabricator.wikimedia.org/T215351 (10Halfak) I was able to find 172 pages that span the predicted quality spectrum. I think labeling them will boost the m... [14:52:47] Technical Advice IRC meeting starting in 10 minutes in channel #wikimedia-tech, hosts: @CFisch_WMDE & @bd808 - all questions welcome, more infos: https://www.mediawiki.org/wiki/Technical_Advice_IRC_Meeting [15:36:29] 10Scoring-platform-team, 10articlequality-modeling, 10artificial-intelligence: Create new labeling campaign for Basque Wikipedia articlequality - https://phabricator.wikimedia.org/T215351 (10Theklan) That sounds great! I think that "Article quality version 2" is not bad... currently I don't think we need som... [15:36:51] halfak: standup? [15:37:16] Oh! I had a meeting scheduled over the top of it. So I'll need to standup from IRC> Sorry for the lack of notice. [15:37:46] I've got the jawiki models tuned and I'm currently working on getting a new article quality dataset loaded into wikilabels for euwiki. [15:37:59] Still blocked on that gerrit mirroring issue for the ORES deploy. [15:39:16] I managed to build a dataset that should let me get started on the threshhold analyis. [15:39:21] I'll also be working on expense reimbursements for past trips and travel approval for upcoming AI ethics policy discussions trip. [15:39:43] I also made some time series plots of the variables [15:40:14] pretty noisy! but when I try to clean them it up it seems like the increase in reverts from the treament to control group looks real [15:42:06] narrowing to reverts that weren't undone within 48h made the regression analysis of that variable more clearly robust, but it's an increase overall, it didn't happen for anon or newcomer edits. [15:43:09] didn't *just" happen for anon or newcomer? Or only happened for non-anon/newcomer? [15:43:24] only happending for all reverts [15:43:32] not for reverted anons or reverted newcomers [15:43:47] * halfak is confused. [15:43:48] maybe narrowing to "undo" reverts will help [15:44:30] If it's happening for all reverts, but not reverted anons/newcomers. Doesn't that mean it's primarily happening for non-anon/newcomer reverts? [15:44:35] yeah [15:44:41] Aha! WEIRD [15:44:55] Catching the damage from the experienced editors O_o [15:44:57] so maybe its spurious? [15:45:22] i'm trying to find more convincing evidence that it has any influence on behavior at all [15:45:38] 10Scoring-platform-team, 10articlequality-modeling, 10artificial-intelligence: Create new labeling campaign for Basque Wikipedia articlequality - https://phabricator.wikimedia.org/T215351 (10Halfak) See https://labels.wmflabs.org/ui/euwiki/. Currently the campaign has a English Langauge name. We can rename... [15:45:39] that's the initial goal of the threshhold approach [15:45:48] Gotcha. [15:46:01] if y is time-to-revert and x is the ores score shouldn't there be jumps at the threshholds? [15:46:23] downward jumps [15:46:27] Assuming people are using RCFilters. [15:46:51] yeah this is trying to verify that people are using RCFilters [15:47:26] Roan said there were some event logs or something for RCFilters [15:47:40] I should take a look at that as well, but I have to find the table [15:47:59] I was also under the impression that they were turned off and so are probably incomplete [15:48:04] 10Scoring-platform-team, 10articlequality-modeling, 10artificial-intelligence: Create new labeling campaign for Basque Wikipedia articlequality - https://phabricator.wikimedia.org/T215351 (10Theklan) I have opened it and the first one is a diff of one letter in an article. What should I do with that? [15:48:04] we should turn them back on! [15:49:53] 10Scoring-platform-team, 10articlequality-modeling, 10artificial-intelligence: Create new labeling campaign for Basque Wikipedia articlequality - https://phabricator.wikimedia.org/T215351 (10Halfak) Looks like I made a mistake. One sec. [15:51:49] 10Scoring-platform-team, 10articlequality-modeling, 10artificial-intelligence: Create new labeling campaign for Basque Wikipedia articlequality - https://phabricator.wikimedia.org/T215351 (10Halfak) OK should be fixed. [16:54:44] 10Scoring-platform-team, 10Wikilabels, 10articlequality-modeling, 10artificial-intelligence: Build article quality model for Dutch Wikipedia - https://phabricator.wikimedia.org/T223782 (10Ciell) In 2006 (I know, 13 years ago already) we voted on this Quality scale and the idea was declined by the Dutch com... [18:42:15] halfak: RoanKattouw I'm taking a look at the event logging from changesfiltering [18:42:54] I want to be clear that the data I found is in the event.changeslistfiltergrouping table [18:43:18] Yes that one is still around [18:43:22] the event.changeslistfilters and event.changeslistgroupings are empty [18:43:23] One of the other ones was disabled [18:43:28] Uh what [18:43:41] The former was disabled, I know that (because I did it) but I thought the latter was still around [18:43:42] sorry [18:43:48] the latter is still around [18:43:53] Maybe it's just because Analytics Eng no longer replicates most things to MySQL? [18:43:54] OK cool [18:44:01] event.changeslisthighlights is gone [18:44:22] so when it was disabled that meant all the data is purged? [18:45:38] we should update metawiki to reflect this [18:46:44] what's in changeslistfiltergrouping? there is no metawiki page for it [18:48:40] Running to lunch. Back in an hour [18:49:06] 10ORES, 10Scoring-platform-team (Current), 10editquality-modeling, 10artificial-intelligence: Look at recent changes filters event log to track usage - https://phabricator.wikimedia.org/T225133 (10Groceryheist) [18:50:41] 10ORES, 10Scoring-platform-team (Current), 10editquality-modeling, 10artificial-intelligence: Find out what tools are used for making reverts on the ores-enabled wikis. - https://phabricator.wikimedia.org/T225134 (10Groceryheist) [19:07:37] RoanKattouw: it seems like these are events where people interact with adding or dropping filters? [19:50:15] 10Scoring-platform-team, 10Diffusion, 10editquality-modeling, 10Release-Engineering-Team (Backlog), 10artificial-intelligence: Gerrit repo scoring/ores/editquality not mirroing - https://phabricator.wikimedia.org/T224996 (10mmodell) Worked on this today with Tyler in our pairing session. @halfak: This s... [20:07:07] 10Scoring-platform-team, 10Diffusion, 10editquality-modeling, 10Release-Engineering-Team (Backlog), 10artificial-intelligence: Gerrit repo scoring/ores/editquality not mirroing - https://phabricator.wikimedia.org/T224996 (10mmodell) [20:21:57] o/ [20:22:19] Ended up doing a couple of meetings from my car :| [20:22:36] Looks like I'm going to start early tomorrow to make up some work. [20:22:56] groceryheist: what wiki was I going to look at again? Was it etwiki? [20:32:32] et wiki has a crazy spike [20:32:44] something to look into [20:32:55] but it isn't near the cutoff [20:33:55] but when you only look at anon reverts there's a clear change [20:34:02] so yeah i think it was etwiki <- halfak [20:34:18] groceryheist, got it. Thanks. [20:34:25] Now to figure out who we worked with on etwiki. [20:34:54] https://phabricator.wikimedia.org/T159608 [20:39:15] https://et.wikipedia.org/wiki/Kasutaja_arutelu:Cumbril#How%27s_ORES_working_out_for_you%3F [20:40:34] ok I'll reach out to Eran in a similar way [20:44:27] I just started a process to extract all reverts from etwiki. [20:44:39] $ mwreverts dump2reverts /mnt/data/xmldatadumps/public/etwiki/20190501/etwiki-20190501-pages-meta-history.xml.bz2 --window 2 --radius 48 > etwiki-20190501-reverts.json [20:45:45] groceryheist, do you have the ORES deploy date for etwiki handy? [20:47:39] 05/09/2017 [20:48:42] Thank you! [21:39:48] OK I'm out of here. I've got some analyses running and I've almost finished my expense report push! [21:39:54] have a good one, folks. [22:02:21] RoanKattouw: do you know the sampling rate for changeslistfiltergrouping? [22:12:04] groceryheist: I believe it's 100%, I can't find any sampling logic [22:12:29] ok [22:12:50] and it is just when people change the filters? not when they use links? [22:13:05] it would have been nice if we had saved the other data [22:19:55] Yes it's only for the new interface and when they change their filters, although the initial state might also be recoreded [22:20:35] ok thanks [22:47:43] also we're only storing this for 90 days, so we don't really get to look at most of the transitions [23:02:51] 10ORES, 10Scoring-platform-team (Current), 10editquality-modeling, 10artificial-intelligence: Look at recent changes filters event log to track usage - https://phabricator.wikimedia.org/T225133 (10Groceryheist) The changeslisthighlights and changeslistfilters schemas were deleted along with the data. So w...