There is a list of a few hundred English misspellings in Microsoft Word's Auto-Correct List file. It's a DBCS file but I haven't bothered to figure out the actual encoding. I just stripped any words that included non-ASCII, then started adding my own observations to the file.

With that as a starting point, I wrote a little HTML proxy called Typoxy. When I use it, the proxy corrects spelling and minor grammatical errors as I browse the web. Very nice way to keep my sanity as I browse certain blogs that are littered with examples of the failings of American schooling. I suppose it could just count the errors instead, and thus get a somewhat slanted statistical view of the world of mistakes.

I think any such research should be amended to the Moby Lexicon Project.