tld;dr: Looking for a way to procure consistent recognition of individual characters.
Looking through the history I know this has been asked before but some time has gone by so I thought maybe there had been some updates. I'm creating a program used in music that needs the user to name a note (A,B, C etc.). My success rate in getting Alexa to understand individual letters is very low. Like others I've improved it by creating arrays of possible misinterpretations of letter and searching through those but the success rate is still not good enough for a live skill. Several people suggested putting period after the letter (c., or d. etc.). That seemed to help a bit but not much.
I see there is still not a built in slot for numbers. I've created a custom slot and listed the letters but again no help. I can repeat the same letter in the same tone of voice and volume over and cover and she will hear D as teenager,diner, die, and denaturalize,etc.
I noticed there is a live (by live I mean public available, I don't know the specific terminology here) called Tone Drone that works on the same premise where you name notes and it seems to procure it moral every time so I know its possible. Just wondering whether anyone has any input on how that would be done.
One thing that I've heard is that the voice recognition in the testing simulators is worse than the live one. Is it possible that a lot of these letter recognition problems are solved when the skill is live?
Source: amazon.com