[14:16:15] <halfak>	 o/ glorian_wd 
[14:19:26] <glorian_wd>	 halfak: I saw your last chat last night
[14:19:37] <glorian_wd>	 "I think you'll want to use the proportion in your calculations. "
[14:21:19] <glorian_wd>	 halfak: so, if I understood you correctly, I came up with an idea of summing the probability for determining the weight. That way, we will not eliminate any property pairs which have low probability. 
[14:21:58] <halfak>	 Maybe you could calculate the score you came up with for a few items in the labeled set. 
[14:21:59] <glorian_wd>	 For example, if an item has 3 property pairs with weight 1, 0.5, 0.2, the weight of this item would be 1.7
[14:22:11] <halfak>	 This might help you see if what you have corresponds to reality. 
[14:22:39] <glorian_wd>	 halfak: oh do you mean, I should try this approach first and evaluate the result?
[14:22:53] <glorian_wd>	 whether the result corresponds to the reality
[14:23:27] <halfak>	 glorian_wd, I think you should divide the sum(probability of property present in the items) by the sum(all potential probabilities)
[14:24:39] <halfak>	 Assuming we have the following probabilities {1: 1, 2: 0.5, 3: 0.2, 4: 0.05, 5: 0.06}
[14:24:54] <halfak>	 If an item has 1,2,3, then it would sum to 1.7
[14:25:07] <halfak>	 Total probability is 1.81
[14:25:27] <halfak>	 So score would be 1.7/1.81 = 0.939
[14:25:32] <glorian_wd>	 Gotcha
[14:25:41] <glorian_wd>	 That would be a good formula
[14:25:44] <glorian_wd>	 for determining weight
[14:27:04] <glorian_wd>	 thanks halfak 
[14:27:48] <halfak>	 :) 
[15:29:50] <halfak>	 Amir1, https://www.mediawiki.org/wiki/Wikimedia_Scoring_Platform_team
[15:29:58] * halfak gets ducks lined up. 
[15:30:08] <halfak>	 http://idioms.thefreedictionary.com/have+ducks+in+a+row
[15:30:10] <Amir1>	 That picture ewww
[15:30:47] <Amir1>	 :D
[15:31:19] <Amir1>	 halfak: Can I change the picture?
[15:31:23] <halfak>	 Sure!@
[15:31:35] <Amir1>	 Overall, it looks great
[15:34:41] <Amir1>	 halfak: https://commons.wikimedia.org/wiki/File:Amir-IMG_4549.jpg
[15:34:47] <Amir1>	 Do you think would be okay?
[15:35:00] <Amir1>	 it needs landscape 
[15:35:01] <halfak>	 Yup :) 
[15:35:27] <halfak>	 Oh!  We should add awight
[15:37:01] <Amir1>	 https://www.mediawiki.org/w/index.php?title=Wikimedia_Scoring_Platform_team&diff=2464372&oldid=2464366
[15:37:22] <halfak>	 lol @ moody teen
[17:13:21] <Amir1>	 Okay, Now, it's ores time :)
[17:13:29] <halfak>	 \o/ 
[17:13:50] <Amir1>	 for today I have this https://phabricator.wikimedia.org/T162620 and gross javascript fixing 
[17:13:56] <halfak>	 Amir1, I got a ping from the collab folks that they'll be making some announcements starting tomorrow and that people will be directed towards our docs. 
[17:14:18] <halfak>	 If you feel the muse, it'd be cool to have you check out :m:ORES and :m:ORES/Get support. 
[17:14:26] <Amir1>	 Sure!
[17:14:27] <halfak>	 Otherwise, I'll spend some time with it today. 
[17:14:48] <Amir1>	 I will do it once I'm done with Bengali
[17:14:52] <Amir1>	 I take a pass
[17:15:09] <halfak>	 Cool. 
[17:15:28] * halfak puts final touches on a blog post about bot fights. 
[17:15:43] <Amir1>	 nice!
[17:15:51] <halfak>	 woops.  Better eat lunch before I don't have time anymore. 
[17:15:53] <halfak>	 back in a bit
[17:54:09] <Amir1>	 halfak: for when you're back
[17:54:10] <Amir1>	 https://meta.wikimedia.org/wiki/Objective_Revision_Evaluation_Service/Support_table
[17:55:48] <Amir1>	 I did some updates but I think some more needs to be made based on http://tools.wmflabs.org/dexbot/tools/wikilabels_stats.php
[17:57:03] <halfak>	 Amir1, thanks.  I'm not sure how I feel about this at the moment.  We should reconsider how we report status so that we don't need to be updating Wiki pages all of the time. 
[17:57:35] <Amir1>	 Yeah, Agreed
[17:57:53] <Amir1>	 maybe we need a CL to do that
[17:58:20] <halfak>	 Seems like that's not in the cards for us. 
[17:58:23] <halfak>	 So, maybe a bot :) 
[18:08:30] <halfak>	 Amir1, FYI: https://github.com/wiki-ai/revscoring/pull/309#issuecomment-300244266
[18:09:07] <Amir1>	 yeah, bot doesn't sound too horible
[18:12:15] <halfak>	 Amir1, you've comfirmed the tests run locally?
[18:13:09] <Amir1>	 halfak: I had to disable bad word and stopwords tests, I don't know why they fail on my localhost
[18:13:23] <halfak>	 Oh!  We'll need to figure that out. 
[18:13:28] <Amir1>	 I added a comment in the test file
[18:14:13] <halfak>	 I'll check it out. 
[18:19:37] <travis-ci>	 wiki-ai/revscoring#963 (thresholds - 8617d7f : halfak): The build was broken. https://travis-ci.org/wiki-ai/revscoring/builds/230461778
[18:24:43] <halfak>	 Ahh!  Word boundaries
[18:28:22] <halfak>	 Amir1, https://github.com/wiki-ai/revscoring/pull/309/commits/3018e46e7bbb37cbfcb0a442444b4cbef51f4b3b
[18:28:45] <halfak>	 Adds bengali signing chars handling to word boundary matching :) 
[18:28:50] <Amir1>	 halfak: you rock
[18:29:03] * halfak flexes :D 
[18:29:35] <halfak>	 {{merged}}
[18:29:35] <AsimovBot>	 10[2] 04https://meta.wikimedia.org/wiki/Template:merged
[18:29:39] <halfak>	 Nice work! 
[18:41:44] <Amir1>	 halfak: new commit on progress report page
[18:41:47] <Amir1>	 it's way cleaner
[18:41:49] <halfak>	 \o/
[18:42:17] <halfak>	 I think I've got all my management and comms duties out of the way.  Time to work on thresholds/stats. 
[18:43:38] <Amir1>	 halfak: before you go, Is there anything else I can do for today?
[18:44:04] * halfak looks and thinks. 
[18:48:09] <halfak>	 Amir1, what do you think about the Tensor Flow testing?
[18:48:25] <halfak>	 Looks like there hasn't been an update there in a while. 
[18:48:26] <Amir1>	 good thing. I was at the middle of it and forgot
[18:48:39] <Amir1>	 no, it has been working on unbalanced sample
[18:48:46] <Amir1>	 but I need to retrain the GA
[18:50:28] <halfak>	 Gotcha.  Does that feel like enough of a task for now?
[18:51:40] <Amir1>	 halfak: it's working for now, I need to wait until tomorrow 
[18:51:53] <halfak>	 Gotcha.  So need something else then, right?
[18:52:01] <Amir1>	 yup :D
[18:52:12] <halfak>	 Oh!  I have one that might be fun. 
[18:52:26] <halfak>	 Implement the basic item quality model for Wikidata
[18:52:52] <halfak>	 Just use a similar set of features that we use in editquality and see how good of fitness you can get. 
[18:52:58] <halfak>	 It should be crappy, but it'll be a good baseline. 
[18:54:00] <Amir1>	 okay, is there a phab card?
[18:54:27] <halfak>	 https://phabricator.wikimedia.org/T157498
[18:54:46] <halfak>	 You might want to make a sub-task for implementing a "basic model"
[18:55:02] <halfak>	 Up to you 
[18:55:10] * glorian_wd peeking at this task
[18:55:14] <Amir1>	 I make a subtask
[18:55:14] <glorian_wd>	 Hey2.. involve me on that task :D
[18:56:05] <halfak>	 glorian_wd, your work should be independent. 
[18:56:21] <halfak>	 Amir1 has the ORES side and you'll have the "get signal for completeness" side. 
[18:56:38] <glorian_wd>	 huft okay
[18:56:46] <halfak>	 Hopefully both will come together soon ;) 
[18:57:20] <glorian_wd>	 Amir1: do consider the quality criteria on https://www.wikidata.org/wiki/Wikidata:Item_quality when you are working on the model
[18:59:37] <Amir1>	 halfak: just a note, please don't forget to review https://github.com/wiki-ai/wikilabels/pull/179
[19:00:56] <halfak>	 {{merged}}
[19:00:56] <AsimovBot>	 10[3] 04https://meta.wikimedia.org/wiki/Template:merged
[19:01:09] <halfak>	 we need to teach AsimovBot to do something better when we do a merged template :D 
[19:01:15] <halfak>	 Like {{done}}
[19:01:15] <AsimovBot>	 You rule, halfak!
[19:01:18] <halfak>	 Thanks AsimovBot 
[19:01:22] <glorian_wd>	 lol
[21:12:33] <Platonides>	 ^ jem