There was a book written by a man with locked-in syndrome after a stroke. It made me want to write a living will.
"The entire book was written by Bauby blinking his left eyelid, which took ten months (four hours a day). A transcriber repeatedly recited a French language frequency-ordered alphabet (E, S, A, R, I, N, T, U, L, etc.), until Bauby blinked to choose the next letter. The book took about 200,000 blinks to write and an average word took approximately two minutes. The book also chronicles everyday events for a person with locked-in syndrome."
I wonder which would actually faster - binary search on the standard ordered alphabet or going linearly in order of letter frequency.
Even better - some hybrid algorithm that goes through the most common letters then goes into binary search for the remaining letters (many of which appear at approximately equal frequency)
You could probably also have a tree of the ~32 most common words. At the very beginning you could have a branch to choose whether you want to pick the 'word' tree or the 'letter' tree.
"The entire book was written by Bauby blinking his left eyelid, which took ten months (four hours a day). A transcriber repeatedly recited a French language frequency-ordered alphabet (E, S, A, R, I, N, T, U, L, etc.), until Bauby blinked to choose the next letter. The book took about 200,000 blinks to write and an average word took approximately two minutes. The book also chronicles everyday events for a person with locked-in syndrome."
http://en.wikipedia.org/wiki/The_Diving_Bell_and_the_Butterf...