After a fairly large number of emails I've started working on my type-ahead, predictive text editor project. In support of this effort I'm looking at different algorithms to best predict the word the user next wants to type. The first part of this is looking at documents I've written in the past, and analyzing the frequency of word occurrences within those documents. (I'll skip the details of my theory here, but it involves looking at how often we re-type words compared to how often we use new dictionary-based words beginning with the same characters.)