English Letter Frequency Counts: Mayzner Revisited or ETAOIN SRHLDCU by Peter Norvig is an analysis of English letter frequencies using the Google Corpus Data. Among other things it contains the frequency of all bigrams.
This gist contains a program that extracts those bigram frequencies into a easily usable JSON format.
It also contains the result of running that program (bigrams.json), as well as a version of it where the order of the letters of a bigram is not taken into account (pairs.json). The two JSON files were generated from a copy of the above article retrieved 2015-08-23.
To regenerate the JSON files:
$ curl http://norvig.com/mayzner.html >article.html
$ npm install
$ node extract <article.html >bigrams.json
$ node bigrams-to-pairs <bigrams.json >pairs.json
All of the files are in the public domain.
Does anyone know where to find the frequencies of the bigrams involving non-alphanumeric characters?