I get that but permuting over typos, letter casing, lengths and combination of words would make the dataset huge.

Is there a massively collaborated rainbow table database that is constantly growing? Are there other heuristics that come into play such as guessing the password length or some such thing?