[19:17:39] Hey folks. We'll be starting the research showcase in 13 minutes. [19:20:36] What's on the docket today halfak? [19:20:53] https://www.mediawiki.org/wiki/Wikimedia_Research/Showcase#February_2017 [19:21:03] Urban/rural divide and Navigation vectors [19:22:46] * almost in progress [19:22:53] t -8 minutes [19:31:36] Here we go [19:31:39] Stream should be up [19:31:45] Isaac is starting to talk now [19:42:50] I forgot to note that you can ping me with questions. [19:42:57] I'll deliver them to the speakers :) [19:44:29] Just say "halfak" and I'll get the ping. [19:44:43] thanks halfak ;) [19:44:57] * halfak receives ping [19:44:58] :) [19:46:31] Woo name droppin' [19:47:43] We like robots in MN [19:51:15] that map reminds me of the translation map we did in our 2012 paper, where Italian Wikipedia had translated lots of articles about Montana and Wyoming :) [19:52:44] FAs on Covered Bridges in Pennsylvania are almost all from one user Ruhrfisch [19:55:26] 5 minute warning [19:55:28] Any questions? [19:56:16] Just a note that Lsjbot have made a lot of bot edits on geography articles on svwiki and cebwiki, might be interesting to look into further [19:57:44] is there a difference between places that are close to urban areas, e.g. 10 miles, 100 miles, 200 miles, population within those areas [19:58:02] halfak ^^ [19:58:23] From YouTube live chat: "Have you seen the geography articles by Lsjbot on svwiki and cebwiki?" [19:58:31] halfak: ^ [19:58:36] Thanks. WIll ask that next [19:58:42] Started talking before i saw it :S [20:00:01] this is very interesting research, but i am wondering whether (eg.) per capita is really the right weight against which to assess the sucess/failure of peer production [20:01:20] halfak, brendan_campbell I responded to jan on Youtube [20:01:20] e.g. it couldbe that local content in urban areas is of more relevance to rural readers than vice versa [20:01:32] J-Mo: right on [20:01:39] think white house vs. village town hall [20:02:15] so we should be careful before drawing too many conclusions about the (certainly existing) biases of wikipedia's process [20:02:24] I just invited Isaac to join us here. Sorry I didn't get to all of the questions. [20:02:49] halfak: no worries about mine, i posted it quite late here anyway [20:03:43] halfak: No worries, it was more of a tip than a question [20:04:54] J-Mo: It is me on youtube live chat too :) [20:05:00] ohai :) [20:05:06] https://sv.wikipedia.org/wiki/Special:Bidrag/Lsjbot [20:05:27] welcome isaacj! [20:05:33] thanks! [20:05:43] anyone who had a question for Isaac that didn't get asked—here he is [20:05:48] :D [20:08:13] brendan_campbell: it looks like there are *three* chat channels now - IRC, hangout, and youtube live? [20:08:21] would be nice to limit fragmentation [20:08:28] isaacj: in your map of article density in China, there’s a yellow area (mid density) in a sea of read (low density), is that a city? [20:08:30] hangouts isn't really active [20:08:43] More in the business of managing the stream [20:08:57] isaacj: Hi, I just wanted to give you a pointer that might be interesting, User:Lsjbot on svwiki and cebwiki. >1 million bot generated geographic articles [20:09:27] one interesting thing that can be done in word2vec is the relationship between things, the semi-famous example being: king - man + woman ~= queen. Any investigation of of similar relationships are found with article2vec ? [20:09:54] interesting question Nettrom - i'll be the first to say i'm not an expert. it does look like there's a city there though. Urumqi [20:10:36] HaeB: I just figured out how to disable YouTube Live chat. I can do so next time if event organizers wish. I find that YouTube Live is a nice alternative to those who are intimidated by IRC. We are able to interact with a wider audience. [20:13:51] Ainali: thanks for the pointer! yeah, the range of human vs. bot-generated content across the different editions is pretty amazing. definitely a lot of open questions about how this looks like in other language editions. one of the things i've been thinking about is how you might make bot generation a little more "local" to the area. [20:14:03] brendan_campbell I agree that having an alt venue for questions is useful. I'd prefer to keep YouTube live chat enabled for these showcases. [20:14:11] it's not too hard to monitor one more thread [20:14:26] thats https://meta.wikimedia.org/wiki/Research:Wikipedia_Navigation_Vectors BTW [20:14:42] http://tools.wmflabs.org/readmore/ [20:14:53] bah! Beat me to it [20:15:55] The most similar articles to "Aaron Halfaker" include "Whoopie Goldberg" and "World War I" [20:16:40] makes sense to me [20:16:43] ;) [20:17:08] Oh wait. It looks like it just showing me the old results. [20:17:31] Looks like "Aaron Halfaker" breaks it. [20:17:54] halfak: http://recommend-related-articles.wmflabs.org/types/related_articles/v1/articles?source=en&seed=Aaron%20Halfaker [20:18:23] Do you think it's just not got enough words? [20:18:26] https://www.mediawiki.org/wiki/Reading/Web/Projects/Related_pages [20:18:44] Breaks for "John T. Riedl" too [20:18:45] or not enough traffic? I know that Ellery sets some thresholds [20:19:03] That'd make sense. [20:19:16] well i know that i'll start each morning now by reading about Aaron Halfaker and then all sorts of conspiracy theories [20:19:45] * halfak wants some juicy conspiracy theories to go along with his bland conspiracies. [20:20:08] Looks like we're wrapping up. Any questions about Navigation vectors? [20:20:15] Aaron gets ~20 views a day, but maybe they don’t read many other articles afterwards? [20:20:35] What does "Seed item is not in the embedding or no neighbors above threshold." mean? [20:21:01] dataset on figshare: https://figshare.com/articles/Wikipedia_Vectors/3146878 [20:21:15] halfak: one interesting thing that can be done in word2vec is the relationship between things, the semi-famous example being: king - man + woman ~= queen. Any investigation of of similar relationships are found with article2vec ? [20:21:34] got it [20:22:57] brendan_campbell, J-Mo: i'm sure there are quite a few people who prefer youtube live over IRC, but vice versa too (e.g. because of privacy concerns) ;) i would put a {{citation needed}} behind "wider" - a lot of people aren't comfortable creating / publicly using YouTube accounts. ... [20:23:23] ... if we want to seriously go into that direction, i would rather recommend exploring Facebook Live - there's a lively community of wikimedians discussion wikimedia topics on FB; the same can't be said about YouTube or Google+ [20:23:28] I tried it on Melodifestivalen 2017 on svwiki that has over 2000 views/day https://tools.wmflabs.org/pageviews/?project=sv.wikipedia.org&platform=all-access&agent=user&range=latest-20&pages=Melodifestivalen_2017 [20:23:46] ...of course that too excludes people, in particular those who don't like to use real names online [20:24:29] HaeB: I'm certainly open to "alternative" alternative channels :) Just so long as there's a non-IRC option. [20:25:30] HaeB: I'm not opposed to FB live, the streaming process is almost identical to YouTube's [20:26:11] Ya'll are messing me up with your meta talk. Hard to scan for questions :P [20:29:24] * brendan_campbell zips lips [20:30:45] :D [20:30:59] This is a good discussion. Controlling for traffic in different languages could be challenging. For instance, Norway sends about 1/2 traffic to English and Norwegian editions. [20:31:20] Maybe the read very different articles in the two editions? [20:32:35] J-Mo, brendan_campbell: in any case, i think communication fragmentation is a big downside, and i'm not convinced the other benefits outweigh it [20:43:13] Haeb: fair. I still believe we should provide a non-IRC option. we're publicizing these showcases through social media, and many interested parties will not be familiar with using IRC (and the webchat interface can be confusing if you've never used IRC). I'd like those folks, if/when they show up, to be able to participate actively. [22:42:46] yuvipanda, annafil and I may have just broken PAWS [22:43:01] we're both getting 502 bad gateway errors now [22:43:55] J-Mo: eating rn i can take a look in about ten mins? [22:44:10] sure! thanks :) [22:59:17] J-Mo: now? [22:59:47] checking, yuvipanda [23:00:35] works for me! just pinged anna [23:05:38] I haven't heard from annafil, but I think we're good yuvipanda. Thank you! [23:06:15] J-Mo: np!