"Did you mean..." Behaviour

[philipp2310]

There is room for improvement for when Alice didn't understand the user.
The problem can be bad ASR (a) or missing NLU training data (b)

a)
spoken "what is the temperature in the kitchen"
ASR: "temperature in the kitchen"
NLU 40% confidence for intent: "increase the temperature in the kitchen"
*new -> Alice: "did you say [please increase the temperature in the kitchen]" - with the replied text coming from the first entry of dialogTemplate for the recognized intent, filled with the recognized slots

b)
spoken "what is the temp in the kitchen"
ASR: "what is the temp in the kitchen"
NLU 40% confidence for intent: "what is the temperature in the kitchen"
*new -> Alice: "did you say [what is the temperature in the kitchen]" - with the replied text coming from the first entry of dialogTemplate for the recognized intent, filled with the recognized slots

A faster way for now would be to echo the captured input (ASR) to the user and ask if it was his input. This might create very strange and unnatural voice output.