inform extract wikipedia move long tail fei wu raphael hoffmann daniel weld wikipedia comprehens sourc qualiti inform kind intern structur relat summari infobox enabl supervis inform extract previou effort extract wikipedia achiev high precis recal popul class articl fail larger number case larg incomplet articl infrequ infobox lead insuffici train data paper present techniqu increas recal wikipedia long tail spars class shrinkag automat learn subsumpt taxonomi retrain techniqu improv train data supplement result extract broader web experi compar design variat show concert techniqu increas recal factor maintain increas precis 