[18:56:52] bmansurov: do you have some time now? [19:11:45] quiddity: can the languages template be added to a section of the page, as opposed to the whole page? [19:12:44] I think the answer is sadly no. ;) [19:12:54] leila, i think you might mean: can only part of a page be 'marked for translation', to which the answer is yes. [19:13:02] context/link? [19:13:55] quiddity: bad example, but an example: suppose I wanted to signal that we have translation for https://meta.wikimedia.org/wiki/Research:Characterizing_Wikipedia_Reader_Behaviour/Robustness_across_languages#Background and that people should feel free to translate to their language? [19:17:13] leila, yeah, that should be possible. Seems to work: https://meta.wikimedia.org/wiki/User:Quiddity_(WMF)/sandbox3#Background [19:17:43] although it leaves the "translate this page" marker at the very top :/ [19:19:01] leila: I'll have some time in about 30 mins. [19:19:52] quiddity: thanks! very helpful. (though I agree that the "Translate this page" is confusing. [19:20:01] bmansurov: ok. let's talk then. here is fine, I think, or voice. [19:25:15] leila: ok, I'll ping you here [19:58:19] * leila lunch|fast [20:02:03] leila: sorry, just back now. let me know when you're around. I'm here. [20:27:43] bmansurov: soooo, here is what I'm thinking re eliciting new editor interest test [20:28:30] bmansurov: I think we should send a reminder to those who have not unsubscribed since or provided feedback. [20:28:47] bmansurov: I'm not sure about the reminder message itself though. [20:28:55] * leila checks the thread to see if there are already suggestions there. [20:29:01] leila: i see [20:30:57] bmansurov: ok. nothing on the messaging in the threads. [20:31:18] leila: I think there was a suggestion on waiting a little longer? [20:31:48] * leila checks the thread again [20:32:24] bmansurov: the only thing I see is from Ramtin saying that sending it earlier than Tuesday is not helpful. nothing else on timing. [20:32:55] leila: ok [20:33:26] bmansurov: Let's go with it today then, if you still can do it. I'll think about the reminder message now and put it in the wiki page. [20:33:47] bmansurov: let's just triple check to make sure we don't send it to those who opted out or said they're not interested. ;) [20:33:55] leila: will the remainder message be the same for all users? [20:34:09] leila: yeah I have the opt-out list ready. [20:34:14] * leila ponders on the message [20:34:42] leila: I also need the list of users who already finished this step so that I don't send messages to those users. [20:37:25] bmansurov: check https://meta.wikimedia.org/wiki/Research:Voice_and_exit_in_a_voluntary_work_environment/Elicit_new_editor_interests#The_matching_email [20:38:24] leila: ok looks good [20:38:55] bmansurov: anything else you need from me? [20:39:03] leila: do you want me to email everyone at once or split the list like before? [20:39:16] bmansurov: email at once would be great. [20:39:32] ok, i'll email them today and update the meta. ;) [20:39:37] THANK you! [20:39:46] yay [20:39:50] * leila prepares to move to a coffee shop. :D [20:40:03] I'll be back in 30min the latest. [22:02:46] miriam: https://twitter.com/internetarchive/status/1037054092416188416 - nice! (already retweeted it from @wikiresearch) [22:03:28] HaeB wow thanks!! [22:06:20] BTW, how did you decide on using the draft topic prediction model for this purpose? (were there other options considered / what tipped the scale in favor of this one) [22:06:50] (asking because we might have a similar application coming up) [22:13:24] HaeB other options were : using the top of the category network or design a new topic model. The draftopic would give us a good tradeoff between scalability, accuracy and generalizability of the taxonomy used for categorization [22:13:38] does it make sense? [22:15:10] scalability: we could process milion of pages; accuracy: research evaluation showed it was a pretty reliable model; generalizability: the Wikiproject taxonomy is not perfect but groups articles into very intuitive catgeories [22:15:27] good topic of conv for the Research Group on Thu :) [22:21:12] miriam: yes, makes sense, thanks! (i was wondering whether a model for articles instead of drafts could use additional features for accuracy - e.g. categories or links - and might be easier to generalize across languages. but looks like this was the best choice here. [22:21:55] ...i also seem to recall some academic research where folks built their own, but would need to search for specifics) [22:24:01] is there documentation beyond https://github.com/wikimedia/drafttopic/tree/master/drafttopic ? (CC awight ) [22:24:50] i guess i should ask that in #wikimedia-ai instead [22:30:03] :) there is a paper from halfak accepted at CSCW , I can forward it to you but probably halfak has the camera ready version.. and I can give you some tips on how to run models on the stats machines if you need :) [22:32:24] I added a line on the agenda for the WMF_Research_Group [22:33:48] it's almost midnight in London so, signing off here, have a lovely rest of the day in California! [23:34:48] quiddity: can it be that the translate label for section doesn't work in meta:research? :D [23:35:00] * leila feels very illiterate