Is there a good way to do this? I am thinking one could (taking English as an example) treat each word as a base-26 number (o.ne, t.wo, t.hree, ...) and divide them by 26 to normalize values between 0 and 1.