Because of the structure of the approximation, the posterior probability for a word is product of word prior and letter posteriors.
