How to Enable ASR During Prompt Playback Without Interruptions

[Maciej Zasadzki ADVICOS]

What solutions do you recommend in VoiceGuide to enable ASR during prompt playback without interruptions? We know there is a parameter StartAsrAfterPlayEnd that either starts ASR during prompt playback or only after it ends, but it causes the prompt to stop prematurely when set to start during playback. Is there a way to capture early responses but still let the prompt finish before ASR fully takes over?

[SupportTeam]

Prompt playback is stopped when speech is detected on the line.

It'd be rare that someone would continue to speak if the system continues to play a sound file, so system stops playing when this 'barge-in' is detected.

If you want the prompt to be played in full, ignoring any interruptions, then just have a play module play it out in full and start the Speech Recognition afterwards - very few people will expect for what they say while a prompt plays to be recognized.

[Maciej Zasadzki ADVICOS]

I think our requirement is slightly different from traditional barge-in behavior.

With standard barge-in, the prompt is interrupted as soon as speech (or sometimes any sound) is detected. In practice, this can cause the prompt to stop prematurely, preventing the caller from hearing important information and disrupting the conversation flow. This is particularly problematic when the interruption is caused by background noise or other non-speech sounds.

On the other hand, if speech recognition only starts after the prompt has finished playing, many callers assume their response has already been captured. We encounter this quite often, especially with simple Yes/No questions. Users frequently start answering during the last part of the prompt because they already anticipate what the bot is asking for.

We also see this when a prompt presents several options. As soon as callers hear the option that matches their intent, they often respond immediately without waiting for the rest of the prompt to finish. If recognition has not started yet, their response is lost, and the caller then waits in silence expecting the system to react.

What we are looking for is the ability to continue playing the prompt while ASR runs in parallel. An even better approach would be to interrupt the prompt only after the ASR engine confirms that a complete utterance has been detected, for example through a "Recognition complete" event.

This would ensure that prompts are not interrupted unnecessarily because of noise or partial speech, while still allowing callers to respond naturally at any point during the prompt. The system would stop playback only when it has successfully captured a complete user response, providing a much more natural and reliable conversational experience.

[SupportTeam]

Such operation is not supported.

Also suspect many callers would interpret the system not reacting as that their speech was not heard, and would end up repeating their answer or deciding that system must not be working correctly.

If there are some instructions that callers must listen to in full then perhaps first announce to callers that a message must be listened to in full, and then immediately go to next module which plays the question that can be interrupted with callers answer.