A couple of weeks ago, I wrote about language evolution and where complexities in languages come from if they seem to go away when times goes by. This is high level, grammatical evolution, taking place over hundreds if not thousands of years ago.
Today it struck me you could study the relationship between languages in different way. The speech accent archive of the George Masson University has a large collection of non english speakers, pronouncing the same paragraph of english texts. The difference is pronciation is probably related to their own language. The people from the archive already mapped the texts on ipa. It would be interesting to write a little program that would find the shortest paths between these fragments and try to order the native speakers based on how their pronunciation differs. Would the resulting map of languages resemble the normal language maps?