combin linguist statist analysi extract relat web document fabian suchanek georgiana ifrim gerhard weikum world wide web endless sourc knowledg natur languag step exploit data automat extract pair semant relat text document pair person birthdat strategi task find text pattern express semant relat gener pattern appli corpu find pair paper show approach profit significantli deep linguist structur surfac text pattern demonstr linguist structur repres machin learn provid theoret analysi pattern match approach show benefit approach extens experi prototyp system sc eila sc 