[13:15:57] o/ [13:18:04] 10ORES, 10Scoring-platform-team, 10artificial-intelligence: Review model performance for ptwiki 'articlequality' and 'draftquality' - https://phabricator.wikimedia.org/T250809 (10Halfak) Oh good question. If you don't set the "type" param, it just calls it a "misclassification". [13:47:45] Hello halfak I was able to do a speed check on the elasticsearch vs wikitext_split tokenizer [13:48:25] Elasticsearch was approximately two times sloer [13:48:54] Hmm. That's surprising. It'll be interesting to put that in front of some of the search engineers and to ask them to think about it with us. [13:48:58] I want to think this is mainly due to the overhead since elasticsearch is an api [13:49:13] haksoat, could you get a little writeup added to the task showing the script and output? [13:49:23] The check was on my own machine though, so further checks will still be done [13:49:27] Right. Do we have to send the tokenizer information with every request. [13:49:42] Just so long as both checks happened on your machine, it should be fair. [13:50:01] Okay halfak I'll add my findings to phabricator if that's fine [13:50:15] Great. Thank you! [14:53:18] 10Scoring-platform-team (Current), 10Discovery-Search, 10Elasticsearch, 10revscoring, 10artificial-intelligence: Improve the performance and quality of tokenization in revscoring - https://phabricator.wikimedia.org/T248480 (10HAKSOAT) I have worked to convert the regex from being Python to Java compatibl... [14:54:30] 10Scoring-platform-team (Current), 10Discovery-Search, 10Elasticsearch, 10revscoring, 10artificial-intelligence: Improve the performance and quality of tokenization in revscoring - https://phabricator.wikimedia.org/T248480 (10HAKSOAT) I am currently reading up different resources on writing efficient reg... [14:57:23] <18VAALOK7> wikimedia/editquality#720 (master - 5bb8cf4 : Aaron Halfaker): The build was broken. https://travis-ci.org/wikimedia/editquality/builds/679071974 [14:57:35] halfak: I've added some comments to phabricator [14:57:41] https://phabricator.wikimedia.org/T248480#6080716 [15:07:35] 10ORES, 10Scoring-platform-team (Current): Remove pylru requirement from ORES - https://phabricator.wikimedia.org/T251003 (10Halfak) [15:07:42] 10ORES, 10Scoring-platform-team (Current): Remove pylru requirement from ORES - https://phabricator.wikimedia.org/T251003 (10Halfak) a:03Halfak https://github.com/wikimedia/ores/pull/342 [15:10:08] wikimedia/ores#1430 (no_cache_default - b08a2ad : halfak): The build failed. https://travis-ci.org/wikimedia/ores/builds/679077383 [15:11:28] 10ORES, 10Scoring-platform-team: Allow browser caching of ORES responses - https://phabricator.wikimedia.org/T251004 (10Halfak) [15:14:51] Nice work haksoat! I like your next step in the meantime too. [15:17:46] Great [15:17:50] Thanks [16:00:38] halAFK, what happened to the reverted models? I don't see one at https://ores.wikimedia.org/v3/scores/ptwiki/ [16:01:43] Helder, we started taking them down when damaging/goodfaith became available. [16:03:33] hm, ok [16:07:52] halfak, I was looking at https://github.com/wikimedia/editquality/pull/199#issuecomment-498283439 [16:07:52] Were you doing some research on the effects of ORES over reverting behaviour? If so what were the results? [16:15:28] groceryheist, ^ [16:16:36] Gist: Results are still in progress. We were trying to figure out if ORES made people more or less aggressive towards newcomers and anons. [16:23:45] got it [16:26:08] halAFK, what about old wikilabel campaigns? E.g.: enwiki still shows three: [16:26:08] https://labels.wmflabs.org/ui/enwiki/ [16:26:08] while ptwiki only shows the current one: [16:26:08] https://labels.wmflabs.org/ui/ptwiki/ [16:26:41] (we had another one in 2015, I think) [16:31:45] https://labels.wmflabs.org/campaigns/ptwiki/?all=true [16:31:48] Helder, ^ [16:33:21] oh, I see [17:00:39] Hi Helder [17:00:47] yeah I'm doing some research into that [17:00:57] do you have any specific questions [17:01:10] but yeah our results still aren't fully settled [17:01:22] getting close though [17:01:26] not really, I just wanted to read a little more about it, if there was any page about it [17:02:14] Just curious :) [17:02:26] I'm pretty confident that there /is/ an effect [17:02:45] that flagging edits makes them more likely to be reverted [17:02:56] especially at the "very likely damaging" level [17:03:26] still trying to work out if I belive this effect hits different editors groups differently [17:03:53] in the next few weeks I think I'll have much more to share [17:04:31] hm, interesting [19:37:30] halAFK, here is an easy one: https://github.com/wikimedia/articlequality/pull/117/files [20:35:05] Merged [20:35:13] Thanks! [21:16:18] halAFK, I've found another one: https://github.com/wikimedia/articlequality/pull/118/files