[17:17:25] anyone got an idea of something enwiki reference worthy to update https://en.wikipedia.org/wiki/Talk:Wikidata#Reception ? [17:19:27] addshore: that “Wikimedia references” part must be outdated IMHO [17:19:40] Lucas_WMDE: it is, see what i just put on the talk page [17:19:45] look at https://grafana.wikimedia.org/dashboard/db/wikidata-datamodel-references?refresh=30m&panelId=5&fullscreen&orgId=1, P248 is still the most common reference property [17:19:56] addshore: no, I mean the 3.7% you put on the talk page :) [17:19:57] The part reading "As of December 2015, according to Wikimedia statistics, half of the information in Wikidata is unsourced. Another 30% is labeled as having come from Wikipedia." is also horribly out of date. [17:19:57] The latest data that can be found on https://grafana.wikimedia.org/dashboard/db/wikidata-datamodel-references shows that 78.7% of statements are referenced, and only 3.7% of the references state they come from Wikipedia or a Wikimedia project. ·addshore· talk to me! 17:16, 13 November 2018 (UTC) [17:20:26] P248 is the most common reference property, and that’s now specific to Wikimedia projects [17:20:37] so there’s no way only 3.7% of references are to Wikimedia projects [17:20:40] as nice as it sounds :/ [17:21:07] interesting, the dashboard might need updating for that one [17:21:11] I don’t know how that dashboard counts Wikimedia references, but I assume it’s outdated [17:21:18] or i did maths wrong :D [17:21:34] 3.7% looks about right for that small yellow bar though [17:21:42] so I don’t think it’s your maths :) [17:21:55] that’s going to be somewhere in analytics/wmde/scripts/ right? [17:22:16] Lucas_WMDE: it is in the toolking analyzer i believe [17:22:41] Lucas_WMDE: https://github.com/wikimedia/analytics-wmde-toolkit-analyzer/blob/master/analyzer/src/main/java/org/wikidata/analyzer/Processor/MetricProcessor.java [17:22:55] Lucas_WMDE: this probably needs updating https://github.com/wikimedia/analytics-wmde-toolkit-analyzer/blob/master/analyzer/src/main/java/org/wikidata/analyzer/Processor/MetricProcessor.java#L345 [17:23:04] and this https://github.com/wikimedia/analytics-wmde-toolkit-analyzer/blob/master/analyzer/src/main/java/org/wikidata/analyzer/Processor/MetricProcessor.java#L57 [17:23:33] reference snak processing is done at v [17:23:33] https://github.com/wikimedia/analytics-wmde-toolkit-analyzer/blob/master/analyzer/src/main/java/org/wikidata/analyzer/Processor/MetricProcessor.java#L407-L446 [17:23:55] “The below list was generated using a SPARQL query.” [17:23:58] that’s so helpful :D [17:24:00] ;) [17:24:15] right, i really need to get ready for squash now... [17:24:25] ok [17:24:59] ah, so any reference snak that doesn’t use one of half a dozen approved properties just isn’t counted [17:25:17] and the property list is at least three years old [17:25:23] that would distort the numbers somewhat, yes [17:26:34] Lucas_WMDE: indeed [17:26:39] I’ll open a task [17:26:45] i guess the idea was to generate it from sparql at some point :) [17:26:49] but we never got around to it [17:27:00] FIXME: there is probably a better way to store, manage and update this.... [17:29:37] :) [17:30:20] (off-topic – I wonder whether (: is the ASCII version of 🙃?) [17:32:14] 🙃🙃🙃🙃🙃🙃🙃 [17:32:21] okay, really leaving now .... [17:37:45] but apparently we have almost twice as many “retrieved” as “reference URL”, I wouldn’t have guessed that [17:38:05] though I can see where it’s coming from – I think a lot of “stated in X database” automated references also include a “retrieved” date IIRC [19:16:28] Lucas_WMDE: indeed, retrieved doesn't just include from wikimedia projects [19:16:48] so usage of the property shouldnt line up with # of wikimedia project references i guess [19:16:59] no, I expected it to line up with “reference URL” [19:17:03] independent of the Wikimedia thing [19:18:07] anyways, I’m going home now – see you tomorrow :) [19:38:06] \o [21:13:04] WikidataFacts: did you mistype when you said P248 is specific to wikimedia projects? [21:13:18] OH [21:13:21] well, not mistype [21:13:23] but misremember [21:13:31] I was thinking of P143, sorry [21:13:46] oh dear, that mistake’s in the phabricator task too, isn’t it [21:18:05] I updated the task, thanks for the correction [22:37:51] Hey there. Is anyone around who can check out this config change? It updates HTTP to HTTPS and may block Readers Web SEO launch in tomorrow's European midday swat window. https://gerrit.wikimedia.org/r/#/c/operations/mediawiki-config/+/473292 [22:41:11] Hi, niedzielski :) [22:41:24] o/ o/ [22:41:29] I don't think a https there will make anything break... [22:41:36] I've just -1ed, sorry [22:41:50] Would this modify concept URIs? [22:42:21] oh, these are URIs [22:42:37] wmgWikibaseClientRepoUrl are already https [22:44:41] abian: it would, at least for the page linked data we're trying to add as I understand it. So, for example, the new "sameAs" field we're including should point to "https://www.wikidata.org/entity/Q42" not HTTP. [22:45:32] I fear this should be discussed deeply :/ [22:45:46] That's right. The client configuration, wmgWikibaseClientRepoUrl, is already HTTPS. [22:46:37] Hey, Lucas_WMDE, maybe you can also comment on this if you're available (?) [22:46:54] yes, just about to :) [22:47:02] Faster than light :D [22:52:34] Meanwhile, niedzielski, do you know why some Wikidata entities aren't indexed by the web search engines? [22:54:07] abian: sorry, I don't. We're currently trying to help search engines model the data present on Wikipedia pages better by describing their contents using schema.org/Article with links to Wikidata. [22:54:33] Ah, okay :) [22:54:35] abian: there are many errors in the metadata of the templates on Wikipedia that need to be fixed too. [22:55:08] It would be interesting that all of the entities were indexed, even in several languages (versions with ?uselang=xx), but they aren't :S [22:55:23] niedzielski: What kind of errors? [22:56:29] There are many. Most of the ones I've looked at are missing fields that are required or recommended. [22:57:23] I'm still coming up to speed on it all. [22:58:36] Thank you for the feedback Lucas_WMDE and abian . I'll share this information with the team. I really appreciate your expert and such prompt advice. We're going a little crazy trying to deliver this important feature and want to make sure it's correct. [22:59:32] Thank you for your work :) [23:06:06] niedzielski: no problem, I hope everything goes well :) [23:08:18] Me too! I think we're just scratching the surface with improving the modeling on the Wikipedia side. [23:08:41] niedzielski: since I’ve just seen this for the first time, I actually have some other comments on the JSON-LD too… where would be a good place to leave them? [23:08:47] they’re not really related to https://phabricator.wikimedia.org/T209352 itself [23:09:51] Lucas_WMDE: if there are issues, please open new tickets. The "epic" for tracking tomorrows launch is https://phabricator.wikimedia.org/T209306. [23:10:23] alright, thanks [23:10:47] I see we have a SEO tag now \o/