What does it take to understand a page visually at the machine level? That is our guiding question and one that we are initially applying to the problem of detecting footnotes in a large corpus of ...