Search code examples
voice-recognitionbnf

How to listen for "any speech" in VoCon Hybrid and capture what was said

Part of my task is to make a system that simulates dictating and sending a text message using Vocon.

I tried using the <...> aka "garbage" rule like this:

#BNF+EMV2.1;
!grammar messageInput;
!start <messageInput>;

<messageInput> : cancel | <...>;

Unfortunately that means I don't get what was said, but rather I get the "<...>" as a result.

How can I make Vocon listen to any speech, but report back to me what was said?

(If the changes need to be done in the C code rather than the speech input grammar, I'm also interested in that)

Solution

  • I asked Nuance tech support 5 days ago and finally a couple of hours ago, here's what they answered:

    What you described (returning any spoken text) is dictation -- not grammar-based speech recognition.

    To perform dictation with Vocon Hybrid, you use the server-based or cloud-based service via the NMSP or HTTP interface.

    Please search for "dictation" in the Vocon Hybrid 4.8 documentation.

    So summarising, it is not possible to do it the way I was intending to; it requires a different mechanism.