How'd you assemble the corpus? Only some of these books are public domain, did you have to buy/license the rest?