Sometimes you need relative frequencies of words/terms and you don’t already have a Perl script for your data. Google experiments are a quick approximation – just search for the sequence of words in quotes and use the number of page results (at the top) as an approximate frequency.
I used this in my PhD proposal – I wanted IDF values for some words, so I used Google page frequency for a few common words like and and took the maximum to be an approximation of the total number of documents. Then I did the same for the words I needed IDF for and came up with reasonable values in my figures in 5 minutes or less.
I also used this years ago to help a friend with learning English but forgot until I saw a number of xkcd comics that use the method (and some BBSpot showdowns). It can also be used for things like estimating conditional probabilities.