Finding near-duplicate web pages: a large-scale evaluation of algorithms