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

http://unimrcp.org/asterisk

http://unimrcp.org/lex

http://unimrcp.org/polly

1.1 Applicable Versions

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

UniMRCP Modules for Asterisk 1.5.0 and above

UniMRCP Lex Plugin 1.0.0 and above

UniMRCP Polly Plugin 1.0.0 and above

2 Generic Speech Recognition API

2.1 Overview

The module res_speech_unimrcp.so provides an implementation of the Generic Speech Recognition API of Asterisk, based on the UniMRCP client library.

2.2 Configuration Steps

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

Check the name of the profile, referenced in the configuration file res-speech-unimrcp.conf, which is located in the configuration directory of Asterisk. Use uni2 for MRCPv2, and uni1 for MRCPv1.

[general]

unimrcp-profile = uni2 ; UniMRCP MRCPv2 Server

; unimrcp-profile = uni1 ; UniMRCP MRCPv1 Server

Open the configuration file unimrcpclient.xml, located in the configuration directory of UniMRCP and specify IP addresses of the client and the server. In the following example, the Asterisk/UniMRCP client is located on 10.0.0.1 and the UniMRCP server is on 10.0.0.2.

<ip>10.0.0.1</ip>

<server-ip>10.0.0.2</server-ip>

</properties>

2.3 Usage Examples

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

The following Asterisk AGI script written in Python demonstrates a basic flow applicable to an ordinary Lex chat bot, which needs to be set up separately through the AWS Console.

#!/usr/bin/python

import urllib2

import xml.etree.ElementTree as ET

import sys

from asterisk.agi import *

agi = AGI()

# Create speech object

agi.appexec('SpeechCreate')

# Activate speech grammar

agi.appexec('SpeechActivateGrammar',"builtin:speech/transcribe")

# Set welcome prompt

prompt = 'Welcome to Amazon Lex. How can I help you?'

# Play prompt and perform recognition in a loop until dialog is complete

run = True

while run == True:

# Play prompt

agi.appexec('MRCPSynth', "%s" % prompt)

# Perform recognition (including intent detection)

agi.appexec('SpeechBackground')

# Check whether results are available

status = agi.get_variable('SPEECH(results)')

agi.verbose("got status %s" % status)

if status == "1":

# Get result

result = agi.get_variable('SPEECH_TEXT(0)')

result = urllib2.unquote(result).decode('utf8')

result = '<instance>' + result + '</instance>'

agi.verbose("got result %s" % result)

# Parse content

root = ET.fromstring(result)

agi.verbose("got root")

# Find message element

message = root.find('message')

if message != None:

agi.verbose("got message %s" % message.text)

prompt = "\\\"%s\\\"" % message.text

# Find dialogstate element

dialogstate = root.find('dialogstate')

if dialogstate != None:

agi.verbose("got dialogstate %s" % dialogstate.text)

if dialogstate.text == 'ReadyForFulfillment':

# Dialog is complete

run = False

# Play final prompt

prompt = 'Thank you. See you next time.'

agi.appexec('MRCPSynth', "%s" % prompt)

# Destroy speech object

agi.appexec('SpeechDestroy')

Assuming the provided Python script is named agi_lex.py and located in the Asterisk agi-bin directory, associate the script to an extension, for example, 701 in the Asterisk dialplan.

exten => 701,1,Answer

exten => 701,n,AGI(agi_lex.py)

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

Note that the script also makes use of the MRCPSynth application in order to synthesize the next prompt played to the caller.

3 Suite of UniMRCP Applications

3.1 Overview

The module app_unimrcp.so provides a suite of speech recognition and synthesis applications for Asterisk.

3.2 Configuration Steps

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

Open the configuration file mrcp.conf, located in the configuration directory of Asterisk, and add two profiles for MRCPv2 and MRCPv1 respectively. In the following example, the Asterisk/UniMRCP client is located on 10.0.0.1 and the UniMRCP server is on 10.0.0.2.

[uni2]

; MRCP settings

version = 2

;

; SIP settings

server-ip = 10.0.0.2

server-port = 8060

; SIP user agent

client-ip = 10.0.0.1

client-port = 25097

sip-transport = udp

;

; RTP factory

rtp-ip = 10.0.0.1

rtp-port-min = 28000

rtp-port-max = 29000

;

; Jitter buffer settings

playout-delay = 50

max-playout-delay = 200

; RTP settings

ptime = 20

codecs = PCMU PCMA L16/96/8000 telephone-event/101/8000

; RTCP settings

rtcp = 0

[uni1]

; MRCP settings

version = 1

;

; RTSP settings

server-ip = 10.0.0.2

server-port = 1554

resource-location = media

speechsynth = speechsynthesizer

speechrecog = speechrecognizer;

;

; RTP factory

rtp-ip = 10.0.0.1

rtp-port-min = 27000

rtp-port-max = 28000

;

; Jitter buffer settings

playout-delay = 50

max-playout-delay = 200

; RTP settings

ptime = 20

codecs = PCMU PCMA L16/96/8000 telephone-event/101/8000

; RTCP settings

rtcp = 0

3.3 Usage Examples

MRCPRecog

The use of this application is currently limited to one interaction only, as each MRCPRecog call is performed in the scope of a new MRCP session. The application needs to be revised to allow consecutive RECOGNIZE requests be placed in the scope of the same session in order to continue the dialog with Lex.

MRCPSynth

Use the application MRCPSynth to speech synthesis.

[app-unimrcp-2]

exten => s,1,Answer

exten => s,2,MRCPSynth(Welcome to Asterisk,l=en-US&p=uni2)

In the Asterisk dialplan, associate the provided sample to an extension, for example, 802.

exten => 802,1,Goto(app-unimrcp-2,s,1)

Place a test call and listen to the synthesized message.

SynthAndRecog

The use of this application is currently limited to one interaction only, as each SynthAndRecog call is performed in the scope of a new MRCP session. The application needs to be revised to allow consecutive RECOGNIZE requests be placed in the scope of the same session in order to continue the dialog with Lex.