In most circumstances, repetition is the recipe for success. One particular presentation I gave five years ago, I had gone through the presentation over twenty times. While I sounded natural, there was nothing natural about the process of practising for it, locking myself in a hotel room for a day to repeat the presentation again and again until it was completely committed to memory.
However, in voice interaction, repeating a phrase over again because a device or app doesn’t understand you is a recipe for failure (of the device/app). Doing this several times will lead to the app/device’s abandonment very quickly.
There are a few reasons for failure in voice pick up:
Poor connectivity. The Internet is not connected and local speech recognition doesn’t work.
Poor sound quality. The DSP doesn’t work well or there’s too much noise. This can affect endpoint detection, recognition, or other issues.
Hardware problems. Typically, this occurs when a mic resource is hogged by another app.
Other voice chain issues. NLU engine fails or the service that’s being actuated by voice is offline.
A recent example (from my adventures with Bixby) is trying to open up the Dropcam app. Bixby failed after three consecutive attempts (first was connectivity, second was endpoint, third was STT result) and I went to looking for the app manually.
Enough of these and it’s back to using the manual method.