You can feed books to an RNN and it learn how to produce text.
What I'm interested in is an algorithm that, given say 20 letters it suggest, say the best 10 options for the next 10 letters.
So for example it begins with "The cat jumped " and then we get various options such as "over the dog". "on the table" and so on.
My initial thoughts are to first use the most likely next letters. Then find the letter which is most uncertain and change this to the second likely next letter. And repeat this process.
(Then I may have another evaluation neural network to assess which is "best" English.)
In other words I want the RNN to "think ahead" at what it's saying - much like a chess playing machine.