Installation of the FreeSWITCH and the UniMRCP server with the Lex and/or Polly plugins is not covered in this document. Visit the corresponding web pages for more information.

https://freeswitch.org/confluence/display/FREESWITCH/mod_unimrcp

http://unimrcp.org/lex

http://unimrcp.org/polly

1.1 Applicable Versions

Instructions provided in this guide are applicable to the following versions.

FreeSWITCH 1.4 and above

UniMRCP Lex Plugin 1.0.0 and above

UniMRCP Polly Plugin 1.0.0 and above

2 UniMRCP Module

2.1 Overview

The module mod_unimrcp.so provides an implementation of the ASR and TTS interfaces of FreeSWITCH, based on the UniMRCP client library.

2.2 Configuration Steps

This section outlines major configuration steps required for use of the module mod_unimrcp.so with the UniMRCP server.

Create a new MRCP profile (or modify an existing one) in the configuration directory mrcp_profiles of FreeSWITCH. In the following example, the FreeSWITCH/UniMRCP client is located on 10.0.0.1 and the UniMRCP server is on 10.0.0.2.

<!-- rtcp bye policies (rtcp must be enabled first)

0 - disable rtcp bye

1 - send rtcp bye at the end of session

2 - send rtcp bye also at the end of each talkspurt (input)

-->

</synthparams>

</recogparams>

</profile>

</include>

2.3 Usage Examples

Speech Synthesis

Use the speak application for synthesis.

Place a test call and listen to the synthesized message.

Chat Bot

Lex is a conversational engine, which typically requires multiple interactions with the caller until a dialog is complete.

The following JavaScript application demonstrates a basic flow applicable to an ordinary Lex chat bot, which needs to be set up separately through the AWS Console. The application also requires the results to be returned in JSON format, which must be configured in umslex.xml.

// Profile to use

var profile = "unimrcp:uni2";

// Compose initial prompt command

var promptCmd = "say: Welcome to Amazon Lex. How can I help you?";

// Compose recognition command

var recogCmd = " detect:" + profile + " {start-input-timers=false,define-grammar=false,no-input-timeout=10000}builtin:speech/transcribe";

// Compose final message

var finalMessage = "Thank you. See you next time";

// Main recognition loop

while (session.ready()) {

var command = promptCmd + recogCmd;

// Play and detect speech

session.execute("play_and_detect_speech", command);

// Retrieve result

var jsonResult = session.getVariable('detect_speech_result');

if (typeof jsonResult != "undefined") {

if (jsonResult.indexOf("Completion-Cause:") == -1) {

// Parse JSON result

var result = JSON.parse(jsonResult);

if (typeof result != "undefined") {

if (typeof result.transcript != "undefined") {

// Compose next prompt command

promptCmd = "say: " + result.message;

}

// Check dialog state

var state = result.dialogstate;

if (typeof state != "undefined") {

if (state == "ReadyForFulfillment") {

break;

}

else {

consoleLog("INFO", "Recognition completed abnormally!\n");

}

else {

consoleLog("INFO", "No result!\n");

break;

}

// Speak the final message

session.speak(profile, "", finalMessage);

Assuming the provided JavaScript application is named lex.js and located in the script directory of FreeSWITCH, use the following dialplan to invoke the script.

Place a test call and follow the prompts until the dialog is complete.