Some time ago I iterated through all the links at http://storage.googleapis.com/books/ngrams/books/datasetsv2.html http://storage.googleapis.com/books/ngrams/bo

Some time ago I iterated through all the links at http://storage.googleapis.com/books/ngrams/books/datasetsv2.html http://storage.googleapis.com/books/ngrams/books/datasetsv2.... to get their Content-Length. All up it's 21TB compressed. Just US English comes in at about 9GB or something though IIRC (unsure, might be completely wrong). With this being said, the data is very, very very raw and unprocessed (contains things like "xxiv_DET", "X25.000_NOUN", "X1E", "X16_NUM" etc, just to give some random examples from the Xs). Would be a lot of work to sanitize it, but you might get some interesting results in the process. So IOW this would be somewhere between "toy" and "interestingness from chaos".