[17:47:42] hi tgr. I've got your email and looking into Hungarian. My question is: should I list you as point of contact for it if feasible and if yes, what account? Staff or volunteer? :) [18:08:12] halfak: I think the discussion with WPMED is getting quite interesting, but unsure about how strongly I should argument for something… [18:08:33] (in meeting, will comment shortly) [18:09:28] no problem :) [19:15:26] Nettrom, sorry forgot and made lunch after meeting. Looking now. [19:15:59] halfak: no worries, you’re allowed to eat ;) [19:17:42] Were's the discussion you're referring to? [19:18:05] Oh I see https://en.wikipedia.org/wiki/Wikipedia_talk:WikiProject_Medicine/Assessment#WikiProject_Medicine_and_importance_ratings [19:18:37] I never got a ping for this post [19:19:00] weeeird. You totally did link to me [19:25:58] yeah, I thought I did… [19:28:36] OK so... I don't think we should model WPMED's quirks. [19:28:44] But I guess we could see how well the model could learn them. [19:29:24] As far as infra to build, I think we should have a generalized model and leave it out of scope to apply WikiProject specific peculiarities. [19:29:49] It should be pretty easy to design a tool that uses importance predictions and a few rules to exclude certain types of articles. [19:29:55] Seems that WPMED's rules are simple. [19:30:19] Everything except stuff in a few will be able to use the generalized prediction. [19:30:31] And the stuff in the few categories is automatically "low" [19:30:52] yeah, and those are kept out of the most recent analysis (and model) [19:31:08] lol @ hippocrates being low importance for medicine. [19:31:14] nice [19:32:10] from what I can tell, we’re pushing up against what might be interpreted as inconsistencies in WPMED’s ratings, for example that their Mid- and Low-importance articles are all over the place as far as views and inlinks are concerned [19:32:30] but I’m not sure I want to challenge them much on that [19:33:11] Nettrom, I think that they have an aspirational version of importance whereas we want to measure an actualized version of importance. [19:33:14] could instead decide that any tools built around this will have to allow feedback so that articles don’t get proposed for rerating [19:33:24] hence the focus on prevalence of a disease over popularity. [19:34:35] Just double-checked and I got none of your pings :/ [19:34:37] WTF [19:34:59] the way I read their importance rating scale, since it refers to “strong interest from non-professionals around the world” both prevalence and readership needs to be considered [19:35:11] 'Melanin is not terribly important when it comes to medicine.' O.O wat [19:35:11] but in practice it seems they lean on the former [19:36:01] regarding journals, do you do anything to treat links in citations differently than in-text links? [19:36:14] ragesoss, not yet [19:36:23] Working on the notion of "organic" links. [19:36:30] Which are bare links that appear in the text. [19:36:44] We've already shown that it's more predictive of importance in a basic sense. [19:36:57] Now comes a coherent strategy for measuring it a lot. [19:37:31] cool. I started a few stubs about journals, and they account for a huge portion of the 'new link' notifications I get on Wikipedia — more and more incoming links all the time — even though they aren't very important articles. [19:38:23] Nav boxes :) [19:38:52] yeah, although I'd say nav boxes offer a lot more signal in many cases. [19:39:27] like, being linked from a relatively common nav template is itself an editorial judgment about relative importance of the topic. [19:39:44] ragesoss, in some cases, yeah, but I think it's way more noise than signal. [19:39:55] E.g. all of the elements have roughly the same number of links. [19:39:57] That's wrong. [19:40:31] yeah, I guess it depends on the scale. [19:40:53] and whether nav links are swamping organic links. [19:40:58] one of the challenges with links is that the placement on the page also changes its importance… that’s why I used the clickstream data in the model [20:41:18] HaeB, where do I find that analysis you did on header frequency? [20:42:02] halfak: https://meta.wikimedia.org/wiki/Research:Investigate_frequency_of_section_titles_in_5_large_Wikipedias you mean? [20:42:08] Nice. Thanks [20:42:29] (zareen did most of the work) [20:43:53] Gotcha. Good to know