[00:56:59] halfak: https://phabricator.wikimedia.org/T110585 [04:41:26] Amir1|ZzZzZ, \o/ [14:08:29] o/ [14:31:29] o/ Alchimista [14:31:39] hi halfak [14:31:50] ORES is in a good spot, so I'm starting work on the models based on the Wiki labels data. [14:31:50] how's the new training system? [14:31:51] :) [14:32:04] great news! [14:32:06] Good Q. Still gathering features. [14:32:40] Just thought you'd like to know. I'll likely have some preliminary results to most here and on the ORES page in the next few hours. [14:33:01] in one of the tickets i've opened regarding huggle, i was opening the possibility of when someone reverts on huggle, it can be transmitted to the training, labeling it [14:34:37] ah, i've missed where it was discussed, it's in the good way [14:35:24] Alchimista, I'd like to incorporate that data, but I'm not sure how we could given that it will be a biased set of observations. [14:35:42] Well... then again, when we do our labeling campaign, we filter out all of the edits that were not reverted. [14:36:06] So if we had labels for *most* reverted edits, then that would dramatically reduce the amount of work we need to do for a new campaign. [14:36:21] Do you think we can extract sufficient information from the edit comments about the reason for the revert? [14:36:25] Or non-revert? [14:39:19] well, huggle could use a post request, with oauth, and ores would label only some usergroups, like sysops [14:40:24] Yeah. That'd be OK, but I'd rather the data be queryable by anyone. It seems that hashtags might be a good option. http://tools.wmflabs.org/hashtags/ [14:42:39] humm.. hastags makes me remember an old idea. i though of some kind of hash classification of warnings, a standard one that could be used by huggle or other tools [14:43:33] There's subst template comments. They work kind of like this. E.g. in https://en.wikipedia.org/wiki/User_talk:AshdownMusic [14:43:40] and could be used in summary, something like "V:" added on the summary [14:43:45] It seems that this could be adapted to the comment [14:43:47] Yeah [14:44:24] if it would have a global usage, it could be used for a lot of things, including classification, without depending on local messages [14:44:35] +1 [14:44:40] Would be much better. [14:48:50] let me check if user scripts can mark tags too [14:57:18] nop, not yet. so some hash included on comment section would be usefull [14:57:43] let me think on it, and i'll open a ticket on phab, this could be usefull for a lot of things [14:59:28] Alchimista, cool thanks :) [15:00:09] When you do open that ticket, make sure to subscribe me. I'll pull in the guy who is working on the hashtag service too. [15:48:20] Anyone around who can look at these PRs in revscoring? [15:48:26] I'd really like to make use of them [15:48:51] ToAruShiroiNeko, ^ [15:48:54] Amir1, ^ [15:49:05] * halfak considers self-merging [15:55:03] o/ White_Cat [15:55:06] o/ White_Cat_ [16:03:26] halfak, so my go-to for writing maths is LaTeX. It occured to me after starting this though that you might want to copy/paste parts of this to a wiki. Does mediawiki have TeX support? [16:03:40] Yes :) [16:03:47] See the tag. [16:04:16] Ok. Just wanted to make sure before I got too far int. [16:04:17] in [16:04:32] aetilley, https://meta.wikimedia.org/wiki/Help:Displaying_a_formula [16:04:37] Gotcha :) [16:06:14] * halfak writes a progress report while we're waiting for feature extraction. [16:16:28] hello halfak [16:16:41] Hey White_Cat_ [16:16:48] Got some time to review pull requests? [16:16:48] I intend to cover our not-weekly report to cover progress since the last [16:16:51] sure [16:16:55] I'm doing that right now [16:17:07] Just about done [16:17:09] brb [16:17:16] I have been calibrating a 3d printer and a cat with food [16:17:37] how do youo calibrate your ferrets btw [16:23:07] https://meta.wikimedia.org/w/index.php?title=Research_talk%3ARevision_scoring_as_a_service&type=revision&diff=13705749&oldid=13531661 [16:23:34] I havent been up to date with reports like before, sorry about that :( [16:23:50] Ferrets get social rewards and punishments. Good ferrets get attention and playtime. Bad ferrets get timeouts. [16:24:31] my cat just wants attention [16:24:36] too much of it even -_- [16:27:19] I will abuse the huggle angle, is that ok? [16:27:30] depends on what you mean [16:27:37] "support us and you'll get huggle among other things" [16:27:57] for wikis without huggle whom really really want it [16:28:11] in a more subtle wording naturally [16:28:20] I will also emphasise this with IEG people [16:28:37] Oh yeah. That sounds good. [16:29:02] White_Cat_, did you get any information from the IEG people re. disbursement and mid-term? [16:29:09] not yet [16:29:16] thats a WIP and what I will do today [16:29:30] yesterday the meeting was cancelled much to my disappointment [16:29:34] so nothing was discussed [16:30:04] Yeah. They really need to cancel meetings in advance. [16:30:18] They canceled that meeting 10 minutes before it was supposed to *end* [16:30:24] -_- [16:30:34] to be fair no one bothered to show up aside from me [16:30:41] and in the past I was the main user [16:30:49] I think it is a good idea but their execution is flawed [16:31:01] Siko marking herself as unavalable clued me in before it started [16:31:08] which is why I didnt hang up on our call [16:32:52] OK. And with that, it's time for me to run. I'll be gathering feature-sets for our work with the WP:Labels data. I think we'll have some modeling results on Monday. [16:33:42] awesome [16:34:03] * halfak runs away. [16:34:06] I am unsure what to do tomorrow aside from my usual [16:34:08] see you [16:34:18] Review my pull requests! [16:35:40] sure :3 [16:36:11] I will review them with extra love this time. :D