It was a seperate step in my mind. And the same sort of functionality might be useful in identifying connections between source documents; if you have a pool of papers that you've collected over time, useful connections between them might go unnoticed. If the program looks at all the documents dogeared for an active paper, it could see what they have in common (on a superficial level, at least) and look for that in other places. Wouldn't often yield tremendous insight, but it could be handy.