[00:28:21] programming "feeling" [00:28:39] is it just probability? [00:29:03] asymptotic probability of some y->x [00:29:56] *for [03:33:02] Hi All, just read the *When the Levee Breaks* paper. Can someone tell me the difference between an undo and a rollback? [04:20:09] Undo is MediaWiki preparing the undoing edit for you and letting you save it, possibly after editing it. Rollback is an automated tool that reverts to an exact same version. [04:21:15] Undo is for 1 edit only and can be attempted even if the edit isn't the last one. Revert is for 1 or more edits by the same user, and they have to be the last [04:22:07] aetilley ^ [04:22:23] * guillom --> bus, afk. [04:33:03] it's uh the difference between a semi-automatic and a fully automatic: the feds won't let you have a fully automatic [05:26:20] haha [05:26:25] thanks everyone [05:26:32] guillom: harej ^ [05:26:38] harej: ^ [19:23:23] halfak: o/ [19:23:33] Hi Amir1! [19:23:55] I finished the code that gets reverts based on the new system [19:24:13] so we can use it in the new reverted model [19:24:29] I need revert_window and revert_radius [19:25:07] and I can give you list of revids and reverted status [19:25:13] halfak: ^ :) [19:26:00] for wikidata [19:26:09] but it's pretty ad hoc [19:26:25] Amir1, revert_window == 48 hours [19:26:31] revert_radius = 3 [19:27:03] great, you wil have all of them very soon [19:27:23] \o/ [19:28:12] oh btw I scored 107/120 in TOEFL [19:37:26] halfak: you will results in 13 hours [19:37:37] *will have [19:38:25] I wanted to work on my PR but it seems a little bit complicated [19:38:35] will be done in tomorrow [19:50:43] is 107/120 good? [19:50:55] Also no worries about the delay. [20:00:00] Yes. That's firmly good (according to a quick googling) :) [20:02:07] :) [20:02:11] thanks [20:02:16] I was afk [20:02:31] No worries. I was randomly AFK too. :| [20:02:40] Or rather *suprise meetings* [20:02:46] Good ones though. [20:03:19] awesome [20:03:52] I want to give this a lower priority since using the new method we'll have far better results [20:04:02] *this PR [20:04:27] it's better to talk about bias detection [20:04:52] tell me when you have time [20:06:29] +1 [20:06:51] So yeah. Bias detection. [20:07:58] Tomorrow is a big US holiday (Thanksgiving) and that will extend into the weekend, so I'll be AFK for a few days. [20:08:16] I'm not positive that I'll be able to make the meeting on Friday. I'll let you guys know. [20:08:43] I think my best chance to have a conversation will be tomorrow morning. [20:08:56] Between 1400 and 1600 UTC. [20:10:20] that would be okay for me [20:10:46] just drop an event in calendar [20:11:12] speaking of thanksgiving: https://www.youtube.com/watch?v=e2zyjbH9zzA [20:11:25] you shoul definitely watch this [20:11:57] snl was better in 2013ish but still some sketches are golden [20:15:14] Ha! [20:16:06] This is great. [21:45:07] halfak: Afternoon [21:45:17] o/ aetilley [21:45:20] * halfak is meeting [21:45:24] but 10% present [21:45:25] So just for my own enrichment, in the blog post project = context right [21:45:26] ? [21:45:37] Second paragraph after How it workds [21:45:38] words [21:45:40] works [21:45:41] Yes. [21:45:45] cool [21:45:47] "wiki" == "context" [21:45:54] right [21:46:06] * YuviPanda is meeting too [23:27:05] Well... I've experimented with dropping user.is_anon from the model. [23:27:23] Oh and user.age. [23:27:36] We're seeing drops of 0.1 AUC generally. [23:30:20] But... the predictions seem to be good. [23:35:44] It will be interesting to take these for a test run. [23:47:00] I'd really like to see if we can get back to the same level of AUC without resorting to user-status. [23:47:10] At least not newcomer/unregistered [23:47:23] I think that we could use advanced rights. [23:59:59] halfak: Which of the three models are features for ClueBot?