[00:05:52] hola [01:17:17] PROBLEM - wikidata.org dispatch lag is higher than 300s on wikidata is CRITICAL: CRITICAL - Socket timeout after 10 seconds [01:22:58] RECOVERY - wikidata.org dispatch lag is higher than 300s on wikidata is OK: HTTP OK: HTTP/1.1 200 OK - 1461 bytes in 0.144 second response time [04:42:27] PROBLEM - wikidata.org dispatch lag is higher than 300s on wikidata is CRITICAL: CRITICAL - Socket timeout after 10 seconds [04:48:09] RECOVERY - wikidata.org dispatch lag is higher than 300s on wikidata is OK: HTTP OK: HTTP/1.1 200 OK - 1464 bytes in 1.411 second response time [10:17:32] Lydia_WMDE: did you make a ticket? [10:17:46] addshore: doing it right now [10:17:49] =] [10:21:29] addshore: done and ccd you [13:28:39] Today is the last day for comments: (╯°□°）╯︵ ┻━┻ https://www.wikidata.org/wiki/Wikidata:Wikimania_2016 [13:29:59] (╯°□°）╯︵ ┻━┻ [13:31:02] (╯°□°）╯︵ ┻━┻ [13:32:58] Freenode today: https://twitter.com/freenodestaff/status/670162459328684032 [15:37:56] hey, is there some simple way to do named entity recognition with wikidata? [15:42:15] use case: i have some trending "short string" in any language, let's say "ኮልን"... now it would be cool if i could somehow do a perfect title search over all wkipedias and just get Q365 back [15:43:41] i didn't find a way to restrict search to just the titles of the all wikipedia articles [15:43:44] ahh, DanielK_WMDE_ there is no index on term_type [15:44:34] you could use https://github.com/dbpedia-spotlight/dbpedia-spotlight/wiki and then use the dbpedia<->wikidata links. But that's not the most direct thing, might be better to build your own index [15:45:32] well, it would be cool if it was live... [15:46:08] and i thought as wikidata essentially centralized all the "names" for things it would maybe be the right point to ask... [15:46:11] ah well that's one more order of difficulty then [15:47:14] it centralize names yes, but that doesn't mean it indexes on that (and not only canonical names are interesting) [15:47:46] sure, but i'd even be happy with perfect canonical name matching as a first attempt ^^ [15:49:09] addshore: that would be a problem ;) [15:49:15] yeh, bah [15:50:09] jorn: maybe just try with https://query.wikidata.org/ then [15:50:14] no idea if it's live [15:50:21] but just use the label predicate [15:50:53] were all article titles initially imported as "label"? [15:51:35] i remember there was some weirdness wrt. titles and labels, just not what... [15:52:28] rom1504: https://query.wikidata.org/#select%20*%20where%20{%0A%20%20%3Fs%20%3Fp%20%22%E1%8A%AE%E1%88%8D%E1%8A%95%22%0A} [15:52:57] https://www.mediawiki.org/wiki/Wikidata_query_service/User_Manual#Label_service [15:54:01] got to use that service thing apparently [15:54:04] yeah, in the above i didn't restrict the language at all, so it should return something, no? [15:55:41] no [15:55:52] you really need to use the SERVICE keyword [15:56:14] i don't need it for "Cologne" [15:56:17] I think it queries some different part of the kb [15:56:39] https://query.wikidata.org/#select%20*%20where%20{%3Fs%20%3Fp%20%22Cologne%22}%0A [15:57:52] jorn: yes, it finds items for which "Cologne" is the category, not the label [15:58:06] (click on wdt:P373 in the results) [15:59:25] ugh... that's pretty inconsistent with standard sparql :-/ [16:00:36] well, it appears that they don't store the label as a normal predicate [16:01:00] but SERVICE is a SPARQL term though, not sure how it works exactly [16:01:15] the thing is that i don't know the language upfront... so i'd have to specify all ^^ [16:01:35] (still reading up on that though) [16:04:10] the examples on https://www.mediawiki.org/wiki/Wikidata_query_service/User_Manual#Label_service show how to get the label, I'm not sure how to use a known label to find the entity [16:04:28] exactly... [16:05:17] but i can't be the only one who'd like to find the entities to given labels without known language, am i? [16:07:48] yeah well, I've been using knowledge bases in natural language processing tasks (with DBpedia in my case but it seems similar here), and found that asking the SPARQL endpoint to do stuff with strings is not the best way / possible, generally [16:08:14] but maybe someone built an entity linking tool for wikidata yet ? idk [16:13:06] oh [16:13:11] there is https://www.wikidata.org/w/index.php?search=&search=%E1%8A%AE%E1%88%8D%E1%8A%95&title=Special%3ASearch&go=Go [16:13:25] so I guess it must be using something [16:13:58] yes, but i thought that's also searching in descriptions, no? [16:16:18] https://www.wikidata.org/w/api.php?action=help&modules=wbsearchentities [16:17:37] that also says the language parameter is required and one of... [16:18:37] doing several queries would work then [16:19:11] https://www.wikidata.org/w/api.php?action=wbsearchentities&search=%E1%8A%AE%E1%88%8D%E1%8A%95&language=su returns nothing though [16:19:46] 420 requests to ask for a single "all"? [16:19:58] ah but it's not a label https://www.wikidata.org/wiki/Q2346558 [16:20:40] and also not the right Cologne ;) [16:20:52] Q365 [16:21:16] ah [16:21:21] still not a label of that [16:21:50] https://www.wikidata.org/wiki/Q365 ... yes, that's why i asked about "title search over all wikipedias" [16:22:15] it seems wikidata knows the titles (at least it shows them)... [16:22:21] ahhhh [16:22:21] well that should be easier [16:22:28] title might be a normal predicate [16:23:49] well, at least it doesn't seem to be query-able via the SPARQL endpoint ... (the {?s ?p "ኮልን" } query should've found it otherwise, no?) [16:23:59] I wish they had used a normal triple representation instead of that fancy inconvenient one [16:24:29] ^^ me too, but well, i don't complain as long as it works somehow... [16:24:53] jorn: ah yeah, but maybe it is possible to get it using that SERVICE thing [16:25:21] just that i can't figure out if it's just me or it just can't do what i'm after ... ;) [16:25:45] but as soon as i use that SERVICE thing i have to specify the language [16:26:31] you have to specify a list of languages, specifying all the languages would do (even if a little silly) [16:26:45] (assuming it's possible to query for a title using that) [16:26:53] hehe ^^ ok, will try [16:28:19] rom1504: thanks a lot for all the help, i'll have to continue this tomorrow though, gotta run for the bus [16:28:41] http://tools.wmflabs.org/autolist/index.php?language=en&project=wikipedia&category=&depth=12&wdq=&pagepile=&statementlist=&run=Run&find=%E1%8A%AE%E1%88%8D%E1%8A%95&find_label=1&mode_manual=or&mode_cat=or&mode_wdq=not&mode_find=or&chunk_size=10000 [16:28:47] np, I'm interested by these things too ;) [16:29:28] ^ is a different api, but it seems to do what you want [16:29:55] cool, that sounds promising... [16:30:05] man, there are just too many different tools and APIs ^^ [16:31:13] but actually on top it says en.wikipedia.org ?!? [16:32:51] hmm where ? there's language=en in the url but I think it's just the language of the page it displays the result [16:33:44] oh first line [16:34:30] if you remove that first line, it still works so... [20:49:58] Lydia_WMDE: uh, I tought we couldn't have the same label and description? but I added the label "Simon Lewis" with the description "peintre" to two items: https://www.wikidata.org/w/index.php?title=Q21457316&action=history https://www.wikidata.org/w/index.php?title=Q21457309&action=history without problem? [22:11:04] Harmonia_Amanda: I don't see why two items with identical labels and descriptions would be a problem from a technical point of view, since items are differentiated by their numbers. Of course from a practical point of view it's a lousy idea :P [22:11:38] harej: well I tought it was forbidden, and yesterday I couldn't do it [22:11:53] we needed to clear the disambig in the description [22:13:29] I know the item creation form prevents you from doing that, possibly to prevent the creation of a duplicate item. But if you go to an item that already exists and change the label and description to something already in use... [22:22:23] Harmonia_Amanda: yeah two items with the same label and description in the same language should not be possible [22:22:46] i will have a look [22:23:09] Lydia_WMDE: i used labellister to add the labels [22:23:16] maybe it's the reason? [22:23:16] ok [22:23:20] it should not make a difference [22:23:24] but who knows