Voice Management Guide

The Voice Management user guide

Back to guide

Voice Recognition stage

1 - Definition of options

Name: The name given to your recognition. This does not have an impact on the rest of the configuration
Step type:
  • GET + resource URL: You can have a manually or dynamically compiled list of words (for example, use the list of Centrex accounts in real time). We are going to use an .xml document, which can be retrieved using the resource URL.

<speech-context id=”perso” speech-complete=”true” scope=”hint” language=”fr-FR”>
<phrase>1</phrase>
<phrase>2</phrase>
<phrase>3</phrase>
</speech-context>

id => ne pas changer
language => langue “fr-FR”
scope => hint (texte contient) / strict (texte exacte)

  • Standard + Grammar: You can copy and paste the entirety of an .xml file into the “Grammar” field, which will allow you to have your grammar(*) directly in the interface.
  • Simple + Grammar: If you do not have an .xml file, but merely a collection of words, you can fill in the Grammar field. You must enter one word per row.
  • Session: This configuration is to be placed before and after a “Recognition” stage group. If you have several sequential recognitions, you can check the “retain the resource” box for each stage. This will allow you to retain the recognition resource throughout your entire shift. However, to prevent you from forgetting, and to prevent this from becoming too time-consuming, you can simply apply an initial “Session” configuration with “retain the resource” checked, and then after your “recognition” stages, apply the “Session” configuration again without checking “Retain resource”.

(*)Grammar: a collection of words to be recognised
Input variables
 Personalised variables, @e164_caller, @e164_service…
Sound played prior to the presentation of the person: Sound played to the caller while waiting for the entry. Example: “If you would like to talk to”
Variable : Corresponds to the name of the variable containing the ID of a specific sound to be played. Example: “Cécile”
Sound played after the presentation of the person: Sound played to the caller after the entry. E.g. “Say yes”
Resource waiting sound : The length of time required for recognition can vary. It is thus possible to set a sound while waiting for connection to the server.

  • Duration: If a time is entered and this time is exceeded, the call will be redirected to the step entered under “Step if time exceeded”.
  • Retain the resource: During a call, several successive recognitions may have been put in place. There may be uncertainty with regard to the availability of the recognition resource. Consequently, you will want to retain this resource for as long as the service continues.
    • Check the box: When there are several successive recognitions.
    • Do not check the box: When there are a few seconds of recognition at the start of the service.
Output variables : Name of the variable to recover the word spoken to direct the call, display the agent etc.
Words considered in error: If the text is recognised (in the grammar) but produces error messages, then the call will be redirected to “step if failure”
Go to step: The step to which the call will be directed in case of:

  • success
  • timeout
  • Failure

2 - Practical application

Here is an example of the implementation of voice recognition:
1- Setting up a welcome message
Put a welcome message on your automatic IVR.
2- Applying “Sound” variables in your voice recognition

If you wish to use the same sound that you recorded in your voice recognition several times, you can first create a variable with the “Variable name” found in the “Recognition” stage and the “Value” that corresponds to the ID of the sound, which you can find in the “Sound” tab.


3- Voice recognition stage

The configuration of this voice recognition stage was of the “Simple” type; two words were listed in the “grammar” field: Saturday and Sunday (1). The “Sound played prior to the presentation of the person” is a question regarding the caller’s preference: “Do you prefer Saturday or Sunday?”
We find the variable (2) that we set up in the previous stage, which will play the configured sound.
The output variable (3) has been configured to be linked with the next stage; filtering. This allows you to distribute calls depending on the caller’s response.
Steps (4) have also been configured in case of failure of the recognition, or if the waiting time is too long. You can, for example, apply a sound indicating a technical problem.


4 – Distribute calls based on the caller’s response

Following the “Voice recognition” stage, you can set a “Filtering” stage to filter calls depending on the caller’s response, and redirect the call to the appropriate contact or the correct message. Configure the filter on “Variable” with the same “Name” assigned during the previous stage for the output variable.


In the filters, we find the two words, “Sunday” and “Saturday” from our grammar. Each redirects to its own step.

  • Table of Contents