[20:43:00] ok then [20:43:02] I am trusted? :D [20:43:05] YuviPanda, gotcha. [20:43:15] OK. We still want the USB off change I made. [20:43:21] ToAruShiroiNeko, no! [20:43:23] :P [20:43:28] halfak: http://bots.wmflabs.org/~wm-bot/logs/%23wikimedia-ai/ [20:43:34] halfak: yeah. but now rebase [20:43:49] Helder, can you look at https://github.com/wiki-ai/revscoring/pull/143 again? [20:44:09] YuviPanda, will do. [20:44:54] that looks good to go halfak [20:45:05] halfak, did you notice https://github.com/wiki-ai/wikilabels/issues/68 ? [20:45:13] YuviPanda, should we merge https://github.com/wiki-ai/ores/pull/74 first? [20:45:30] can it simply message and not notice? :/ [20:45:37] halfak: yes [20:45:45] {{merged}} [20:45:53] Helder, campaign isn't done. It turns out that we have double-labels for some tasks due to our auto-labeling. [20:46:15] cool! we have moar bots on this channel! [20:46:18] Helder, see https://meta.wikimedia.org/wiki/Research_talk:Revision_scoring_as_a_service/Work_log/2015-07-22 [20:46:25] yes one bot a day keeps the doctor away [20:46:39] bots distract doctors [20:47:10] It looks like ptwiki will need to go 285+20000 labels to get all the tasks labeled. [20:48:01] halfak: we need to migrate wikilabels to the new setup at some point [20:48:02] Helder, we need to add a 'labeled_tasks' field to the stats [20:48:07] YuviPanda, +1 [20:48:11] halfak: ideally I'd be 100% uninvolved in that and it'll be all madhuvishy and you [20:48:19] OK. [20:48:25] !log set up GitHub notifications for all wiki-ai repos [20:48:34] I'm due for some work on wikilabels to clean up some of the mess we made at wikimania [20:48:36] halfak: not because I want to be uninvolved, but just so that we diversify our investments :) [20:48:46] +1 good plan YuviPanda [20:48:51] You better not be leaving me. [20:48:52] ;) [20:49:02] hehe :D [20:49:04] don't worry [20:52:44] hi hi sup [20:52:51] can/should this bot also post on IRC when someone comments somewhere? [20:53:23] e.g. our main (talk) pages, and maybe GitHub/Phabricator issues [20:53:54] Seems like a fine idea to me. [20:54:07] I don't know if there's integration available for GH comments, but Phabricator is easy, the configuration is in https://git.wikimedia.org/git/labs/tools/wikibugs2.git [20:54:29] Let's not go too far with logging all the things in the channel though. [20:54:49] awight, re. https://github.com/wiki-ai/revscoring/pull/141 [20:54:52] People will begin to suspect we have bots writing bots [20:55:09] yes [20:55:14] Do you want to do the language codes as module names or are you thinking of moving towards Language.get("lang code")? [20:55:44] I commented here, https://github.com/wiki-ai/revscoring/issues/138 [20:56:06] Oh yeah. I saw that. [20:56:13] I'm perfectly happy with all the good stuff happing in PR#141, just looking at future improvements [20:56:24] Makes sense. :) [20:56:29] I did notice that we have a ton of repetition whenever multiple languages are involved [20:56:35] The Makefile, for example... [20:56:49] me too [20:56:57] The actual languages/LANGNAME.py modules are fragile too, cos of that repeated code you were mentioning [20:57:39] We certainly don't have to get declarative if you don't want to, I did see that there's some yamlconfig stuff happening in ores that should be in scope if we ever do decide to clean that stuff up. [20:57:57] Yeah. I'd be down for blocking the pull request until we can get a Regex/Enchant base class implemented. [20:58:15] I don't care about language.get('en') on a personal level, so you should ignore me. It's more that I recognize this as an unsustainable pattern. [20:58:30] Indeed. [20:58:36] It should be *really* easy to create new languages, should not require messing around in more than one place. [20:58:54] Agreed. [20:59:15] A declarative file is a natural fit mostly because it can be read by multiple utilities without having to coordinate the data structures and (programming) languages [20:59:35] I don't want to block your PR, in fact would like to see it merged before any more work happens in that area... [20:59:49] It certainly doesn't make anything worse, all a good idea. [20:59:55] OK. Let's do that. [21:00:04] I'll keep the generalized language class on the backburner. [21:00:11] And run off to my next meeting [21:00:12] o/ [21:00:17] * halfak runs off to next meeting [21:00:21] Unfortunate that you needed module magick, we can clean that up later though [21:00:24] kbye! [21:01:47] Aaron in a meeting? Did that surprise anyone? :D [21:08:28] halfak, what is "labels" in https://labels.wmflabs.org/campaigns/ptwiki/?campaigns=stats ? [21:08:43] Isn't it the number of labels already submitted by users? [21:08:52] Includes auto-labels [21:09:03] In some cases, we auto-labeled things that already had a human label too [21:09:10] So some tasks have two labels! [21:09:53] but 20290 is already greater than the predicted 285+20000 [21:11:08] halfak, ^ [21:55:56] https://labels.wmflabs.org/campaigns/fawiki/?campaigns=stats [21:56:06] halfak: and ToAruShiroiNeko [21:56:13] ToAruShiroiNeko: Simple is ready [21:56:16] Yes [21:56:20] I saw [21:56:21] thanks :) [21:56:26] working on Dutch I hope you won't need it [21:56:43] its non urgent but sure it can wait [21:56:50] this time I am talking to community first [21:56:54] italian is of interest too [21:57:03] because belissimo. [21:58:12] let me pull the phabricator entry [21:58:32] https://phabricator.wikimedia.org/T106861 [21:58:40] https://meta.wikimedia.org/wiki/Research_talk:Revision_scoring_as_a_service/Work_log/2015-07-22 [21:59:07] fawiki has to go 261 [21:59:09] extra [21:59:24] ptwiki only 4 so it should be done really [23:12:20] Helder, indeed it is. looking into it. [23:23:26] Helder, I just requested and received the last 4 tasks in ptwiki :) [23:23:36] So, nothing seems to be broken. [23:25:09] OK. Done! [23:25:23] Now I'll want to have you review the tasks I did with google translate to see if I did OK. :) [23:25:31] Also, mad props for getting ptwiki done! [23:25:33] WOOO! [23:48:07] how can I review them, halfak ? [23:51:38] by any chance are they the ones which appear in my last workset? (coincidentally there are 4 tasks to finish it) [23:56:26] it seems they were the last ones. [23:56:28] I got "conflict: No tasks available for user_id=8303885 in campaign_id=7" [23:56:44] halfak, is this the expected behavior for finishing a campaign?