One potential route to flagging fake news at scale: Linguistic analysis