[19:32:29] <dbb>	 re- the deepLearning citation checker.. this appears to be a related blog post  
[19:32:31] <dbb>	 https://tech.facebook.com/artificial-intelligence/2022/07/how-ai-could-help-make-wikipedia-entries-more-accurate/
[19:56:48] <dbb>	 .. to be blunt, I am sceptical of the Facebook research culture with slick promotion values and glad-handing statements to start.. unfortunately for me, their science is quite good and not dismissable
[19:57:39] <dbb>	 that long post above is new to me, but I did read it once just now.. many particulars
[19:58:32] <dbb>	 maybe there is a place for public-facing posts more of what WMF labs and whatever, is doing
[20:03:22] <dbb>	 one plank to note there is - a previous repo-project is KILT, and that uses a wikipedia dump file. This series above is newer, called SIDE, and has a not-wikipedia web facts text corpus, published open-open-open they say
[20:04:19] <dbb>	 the not-wikipedia training dump is called SPHERE it appears.. I am just reading these parts once.. not an expert, might contain errors by me
[20:06:32] <dbb>	 .. as one example from that post above, the authors claim to get a blog post by a violinist, that references the appointment of a new symphony director, and that replaces a citation in actual wikipedia which was a Press Release (low accountability rank?) 
[20:07:00] <dbb>	 .. thereby strictly adhering to the Wikipedia references rules, and "beating the human editor" too, in an automated way
[20:08:02] <dbb>	 the blog post from a violinist, is inside the SPHERE web facts corpus, it seems
[21:04:38] <dbb>	 .. they ref this page specifically  https://meta.wikimedia.org/wiki/Research:Machine_Learning_Assisted_Wikipedia_Editing
[21:07:50] <dbb>	 .. in talk   MPinchuk (WMF) (talk) 16:57, 12 July 2022