v-LEARNING

Using Voice for Distant Learning in Emerging Regions

Thushari Atapattu and Kasun De Zoysa

University of Colombo, School of Computing, 35, Reid Avenue, Colombo 7, Sri Lanka

Keywords: WWTW, VoiceXML, Transcoder, Asterisk.

Abstract: At present, accessing the internet through visual interfaces is the most common approach. However, it

requires some basic resources such as a computer or web-enabled mobile device, an internet connection,

electricity and some amount of IT literacy. Because of the relatively high cost of this set up, underprivileged

users are unaware or have no direct access to the internet. Since voice communications through telephony

systems do belong to the growing trend, people make use of telephones for various purposes. Among them,

accessing web through telephone devices is explored. In this paper, we present an approach to access the

learning materials of the Learning Management System (LMS) of University of Colombo School of

Computing, Sri Lanka through interactive voice driven applications.

1 INTRODUCTION

World Wide Web (WWW) has become the major

information source around the world. People access

the web for various purposes such as learning,

communication, entertainment etc. As a result, web

has grown to be one of the most popular media in

the world. But, in order to access the internet, it

requires essential resources. Basically, it needs a

computer (or modern mobile device, internet kiosks

etc.), an internet connection (broadband, wire-less,

dial-up etc.), a telephone, electricity and some

fundamental IT skills.

Buying a computer and obtaining a fixed internet

connection might cost around LKR 80,000

(US$800), which is unaffordable for many people in

developing countries. At the same time, such

internet connections are not available in the rural

areas. Majority of the people in developing countries

do not even have electricity in their homes. Due to

lack of resources, it is an overhead for a majority of

people in the world to access internet directly.

According to the internet world stat (Internet

world stat, 2008), approximately 22% of the people

in the world have access to the internet (Table 1).

The table 1 shows that most of the emerging regions

such as Africa and Asia have limited access to the

web. It implies that most of the people in the world

(78%) are still untouched to these sophisticated

facilities.

Table 1: Internet Usage and world population statistics for

June 30, 2008.

World Region Population % Population

(Penetration)

% Usage

Growth

2000-2008

Africa 955,206,348 5.3% 1031.2%

Asia 3,776,181,949 15.3% 406.1%

Europe 800,401,065 48.1% 266%

North America 337,167,248 73.6% 129.6%

Middle East 197,090,443 21.3% 1176.8%

Latin

America/Caribb

ean

576,091,673 24.1% 669.3%

Oceania/Austral

33,981,562 59.5% 165.1%

World Total 6,676,120,288 21.9% 305.5%

According to the statistics, internet penetration is

around 2.2% in Sri Lanka, 2007 (Sri Lanka Internet

world stat, 2008). It is relatively a low rate with

compared to other educational facilities in Sri

Lanka. The situation is raised due several reasons.

The most critical issue is the unavailability of

electricity and internet connectivity facilities in rural

areas. In order to improve these factors, an

infrastructural development should be done which

requires a huge investment.

217

Atapattu T. and De Zoysa K. (2009).

v-LEARNING - Using Voice for Distant Learning in Emerging Regions.

In Proceedings of the First International Conference on Computer Supported Education, pages 216-221

DOI: 10.5220/0001971702160221

 SciTePress

As an alternative, government and private

sectors have invested on public internet accessible

places such as Internet cafes, Public Internet Kiosk

etc. Some of these solutions are still not possible for

underprivileged users as these solutions are also

costly.

Table 2: Internet usage and population statistics in Sri

Lanka.

Year User Population %penetration

2000 121,500 19,630,230 0.5%

2007 428,000 19,796,874 2.2%

Since mobile technologies have been rapidly

growing, people make use of their mobile devices to

access the web anywhere in the world. This will

reduce the overhead of buying a PC with an internet

connection and the expenditure for electricity.

Accessing wire-less networks also requires high

charges, which is not affordable for the people in

developing regions.

Apart from that, accessing internet through

mobile devices also has some additional limitations.

Firstly, buying a featured phone is not less expensive

in Sri Lanka. At the same time, the GSM/GPRS

coverage does not exist in rural areas. The mobile

device is not always user friendly since it has a small

screen and a tiny keypad. All of these limitations

will distract people accessing the web through their

hand-held devices.

All the above mentioned techniques require

browsing the internet through a visual interface such

as a web browser. Due to the above mentioned

limitations, some researchers have explored the

possibility of accessing the web through voice

communication. The basic voice communication has

had a larger penetration among the world population

as well as in Sri Lanka. Therefore, IBM Research

Laboratory (Kumar, 2007) has conducted a research,

which uses voice to access the internet. This concept

is called World Wide Telecom Web (WWTW)

(Kumar, 2007). In this model, the voice sites are

developed instead of typical web sites. Those voice

sites are implemented using a language called

VoiceXML (VoiceXML, 2008). VoiceXML is a

markup language derived from XML. Users are

allowed to call to the voice site which is a collection

of VoiceXML pages.

The preliminary attempt of this work is to build

an interactive voice learning environment for the

undergraduates of University of Colombo School of

Computing (UCSC). Since the cost of basic voice

communication through telephone is relatively low,

accessing web using voice is encouraged. This will

be beneficial for underprivileged students who have

no direct access to the teaching and learning

materials in the web.

This paper is organized as follows. In section 2,

the work related to World Wide Telecom Web is

discussed. Our proposed architecture and overview

of the system is detailed in section 3. The system

functionalities are explained in section 4. Finally the

proposed system is summarized in section 5.

2 RELATED WORK

WWTW (Kumar, 2007) is a tremendous concept of

IBM India Research Laboratory, where voice-driven

eco systems are developed parallel to that of the

WWW. The approach enables deprived population

to become a part of the networked world through

low cost voice communication. This concept was the

basement for various researches related to voice-

enabled applications.

Interactive Voice Response (IVR) systems are

currently most widely used voice-driven applications

in the world. Air-line, hotel reservations, telecom

service providers commonly use these fixed menu-

driven, user input (DTMF) based applications. These

automated systems require high investments and it is

not supportable for non-profit organizations and the

government education sector.

Researchers have developed a low cost IVR by

integrating the existing open source applications and

tools (King, 2006). This system is a hybrid of

OpenVXI (Carter, 2002) and Asterisk (Asterisk,

2008). OpenVXI (Carter, 2002) is a VoiceXML

interpreter developed by speech group at CMU. It

provides APIs for speech synthesis, speech

recognition and telephony services. Asterisk

(Asterisk, 2008) is the mostly used opensource PBX

system in non-commercial applications and

Voiceone (VoiceOne, 2008) is the web based GUI

for Asterisk PBX. The gateway can be utilized to

replace the existing high cost IVR systems.

VOIGEN (Kumar, 2007) enables telephone

subscribers to access voice-driven systems through

ordinary telephone lines. It permits individuals to

create, host and deploy customized voice driven

services. VOISERV (Kumar, 2007) is similar to

VOIGEN (Kumar, 2007) where VOIGEN (Kumar,

2007) create and deliver data services and

VOISERV (Kumar, 2007) delivers converged

services. Both the systems create their own

customized voicesites.

CSEDU 2009 - International Conference on Computer Supported Education

218

The IBM WebSphere Transcoding Publisher

(WTP) (Lamb, 2008) is a commercially available

product that can be used to convert HTML to

VoiceXML. A group from Virginia Tech, VA has

conducted a research to transcode HTML to

VoiceXML using annotations (Shao, 2003).

3 PROPOSED SYSTEM

In order to experiment voice based solutions for

distant learning, we have proposed an interactive

voice driven system which is explained in this

section. Our proposed approach will be developed in

3 stages as listed below,

1. A voice component which gives access to

practice quizes in the Learning

Management System (LMS)

2. Voice site parallel to existing UCSC

Learning Management System

(http://www.ucsc.cmb.ac.lk/lms)

3. Voice module for open source moodle

project (http://www.moodle.org)

At present, we are in the process of developing

the first stage of the system. In order to provide

voice based access to practice quizzes in the LMS,

we have implemented a simple automated Moodle

XML (MoodleXML, 2008) to the VoiceXML

(VoiceXML, 2008) converter. Moodle XML

(MoodleXML, 2008) is a XML based language

which follows XML standards. The quizes of the

Learning Management System could be exported as

Moodle XML. Our converter simply converts the

Moodle XML files to VoiceXML files. The

converted VoiceXML files are intended to interpret

through VoiceXML Interpreter (Carter, 2002).

In the second stage of our proposed project it is

expected to build a voice site in parallel to the

existing UCSC LMS. This would be fully automated

system generated from web system. The voice site is

intended to be updated automatically with respect to

the web system. At the final stage of the project, we

have proposed to build a voice module for the open

source moodle project. This would be beneficial to

the society, as the people are used to customize

moodle for their learning and teaching purposes.

The main focus of our proposed system is to

allow voice access to learning materials for the

UCSC undergraduate, external and postgraduate

students. The system can be sub divided into three

main components.

1. Private Branch Exchange (Asterisk server

and soft phone)

2. VoiceXML Transcoder

3. VoiceXML Interpreter

The Figure 1 depicts the overall architecture of

the system and each of the above sub components

will be discussed in sub sections.

Figure 1: Overall architecture of the system.

3.1 Private Branch Exchange (PBX)

Private Branch Exchange is a telephone exchange,

which serves a particular set of people. It could be

located in a company, school, university etc. The

cost of deploying a commercial PBX system is very

high. Accordingly, we have used an open source

PBX engine called Asterisk server (Asterisk, 2008)

for our project.

Besides, one of the latest trends in PBX

development is the Voice Over IP (VOIP) PBX,

where internet protocols are used to communicate.

The initial focus of the development is to configure a

SIP phone to connect with the Asterisk server. For

this purpose, we have used the freely available Ekiga

(Ekiga, 2008) soft phone. A typical PBX set up is

shown in Figure 2 below.

The asterisk server is basically capable to,

1. Get the user’s input

2. Interactively provide voice response

3. Call forwarding to voice sites

VOIP/SIP phone

Cal

Voiceone

Asterisk

PBX

Web browser

View

Web site

Html

Voice site

Vxml file1

VoiceXML interpreter

TTS

VoiceXML Transcoder

HTML

parser

Rule

engine

v-LEARNING - Using Voice for Distant Learning in Emerging Regions

219

Figure 2: Overview of Private Branch Exchange.

3.2 VoiceXML Transcoder

It is a known fact that web pages are implemented

using HTML. Likewise, voice pages have been built

using a language called VoiceXML (VoiceXML,

2008). As the HTML pages are interpreted visually

through web browsers, VoiceXML files are

interpreted using voice browsers. For that the system

should generate voice pages or convert existing web

pages to voice pages.

The main objective of the system is to

implement a voice site in parallel to the existing

UCSC LMS web site. In order to do that, selected

HTML web pages from the LMS site should be

converted to voice pages. This process could be

done through “HTML to VoiceXML Transcoder”.

As there are no any open source VoiceXML

transcoders available, our system is expects to

implement a VoiceXML transcoder from the scratch.

Our proposed transcoder has 3 main components.

1. HTML parser

2. VoiceXML translator

3. Rule engine

The overview of the proposed transcoder is

shown in Figure 3.

Firstly, the static HTML pages are analyzed

through a HTML parser and a HTML node tree will

be generated. Once the structure of HTML node tree

is analyzed, the page is converted into a VoiceXML

file internally by the system.

When applying the transcoding logic, our system

makes use of grammar rules which have been

defined by us. After validating the conversion with

the rule engine, the syntactically correct VoiceXML

file will be created.

Figure 3: Overview of HTML to VoiceXML Transcoder.

A simple HTML file and it’s corresponding

VoiceXML file is shown in Figure 4 below.

Figure 4: Simple HTML and VoiceXML file.

Internet/ PSTN

Asterisk PBX

Hard

phone

SIP

phone

VOIP

phone

<html>

<head>

<title>Welcome to University of Colombo School of

Computing</title>

</head>

<body>

<form next=”method”

<INPUT type=”radio” name=”degree” value=”computer

science”>computer science<br>

<INPUT type=”radio” name=”degree”

value=”ICT”>ICT<br>

</form>

</body>

</html>

<?xml version=”1.0”>

<form>

<prompt>Welcome to University of Colombo School of

Computing</prompt>

<prompt>Select your degree<enumerate/>

</prompt>

<option dtmf=”1” VALUE=”computer science”>computer

science</option>

</field>

</form>

</vxml>

Transcoding

logic

HTML file

VoiceXML

file

HTML parser

Internal

grammar

rules

External

grammar

rules

HTML node tree

CSEDU 2009 - International Conference on Computer Supported Education

220

3.3 VoiceXML Interpreter

XML based languages require an interpreter to

interpret the markup commands. Accordingly,

VoiceXML files should be interpreted automatically

after the file is altered. OpenVXI (Carter, 2002) is

one of the freely available VoiceXML interpreter

used by majority of voice application builders.

Typical VoiceXML interpreter consists of 3 sub

components.

1. Text-to-speech system (TTS)

2. Voice recognition system

3. User action handler

3.3.1 Text-to-Speech System (TTS)

Text-to-speech system is a way to present text

output to the user through voice communication. In

our system, we are using an open source TTS called

FreeTTS (FreeTTS, 2008). It extracts the output

from the VoiceXML file and presents it to the user

through a soft phone.

3.3.2 Speech Recognition System

A typical voice-driven application has a component

to recognize user's speaking context. In our proposed

system, we have omitted this component and instead

we are collecting user's input through Dual-Tone-

Multi-Frequency (DTMF). The system prompts

choices for the user and based on these choices, user

has to select a number which can be entered through

a telephone dial pad.

3.3.3 User Action Handler

This component is capable of collecting user's input

and respond accordingly. For instance, if the user

does not perform any action at his turn, the

interpreter gives him a second chance to try the

commands or inform him to end the call. Moreover,

user action handler collects user inputs which are

given by the dial pad. Likewise user action handler

automatically performs several intermediate actions

like a human being.

4 SYSTEM FUNCTIONALITIES

In this section, we describe the main functionalities

of the system. The proposed approach is intended to

be accessible via interactive telephone

communication only. The user should make a call to

the system in order to access the contents. The

system is automated to provide services to the user

regardless of other matters. The functionalities of the

system can be categorized into 2 subsections as

follows,

1. User-level functionalities

2. System-level functionalities

4.1 User-level Functionalities

In order to get the benefits from the distant learning

project, the user should place a call to the system.

This can be done through the dedicated telephone

number which is assigned to the voice site. The

user’s call would then automatically be handled by

the Asterisk (Asterisk, 2008) server, where voiceone

(VoiceOne, 2008) is the front end of the server.

The system identifies the call and redirects it to

an appropriate voice site (At present, we have only

one voice site in our system). The system prompts

information to the user and gets their inputs through

DTMF.

A sample user-system interaction is given below,

• User places a call to the system through the

voice number given.

• System: Welcome to Learning

Management System of University of

Colombo School of Computing. Main

Menu, For site news press 1, For

undergraduate courses press 2, For

Examinations press 3, For inquiries press 4,

To exit from the system just Hang-up etc.

• User enters 3 through DTMF

• System: You have selected examinations.

For Time table press 1, For exam results

press 2. To go to the main menu press 0 etc.

The user can navigate through sub menus for his

destination or simply can exit from any menu or sub

menus. If the user fails to respond to the system

within a given time frame, the menu (or sub menu)

will be repeated once. If the user does not respond to

the system further, the conversation will be

disconnected automatically.

4.2 System-level Functionalities

At the system level, VoiceXML files will be

generated and updated dynamically. This could be

done by converting existing HTML files. The

collection of VoiceXML files is integrated as

Voicesites. The VoiceXML interpreter then

interprets these VoiceXML files and presents them

to the user through the TTS according to their

requests. Before the voice prompts are presented to

v-LEARNING - Using Voice for Distant Learning in Emerging Regions

221

the user, VoiceXML files will be validated through

the system.

5 CONCLUSIONS

The V-learning project is proposed for the

underprivileged users to provide access for learning

resources through the voice communication. In our

approach, we have explored the concept of World

Wide Telecom Web that would be parallel to that of

the World Wide Web. The motivation of our

approach is to deliver the services for the benefit of

the students in developing economies. Though it has

several benefits such as low cost, it would not be as

attractive as graphical user interfaces. We believe

that the system would be a bridge between the IT-

savvy and the non-IT-savvy population in the world.

ACKNOWLEDGEMENTS

The author is grateful to the UCSC e-learning centre

for funding the V-Learning project and the WASN

laboratory for providing resources to conduct the

research.

REFERENCES

King, A., Terzoli, A., Clayton, P., 2006. Creating a low

cost VoiceXML Gateway to replace IVR systems for

rapid deployment of voice applications. In

proceedings of the Southern Africa

Telecommunication Networks and Applications

conference.

Kumar, A., Rajput, N., Chakraborty, D., Agarwal, S. K.,

Nanavati, A. A., 2007. WWTW: The World Wide

Telecom Web. In Proceedings of the 2007 workshop

on Network systems for developing regions, Kyoto,

Japan.

Carter, J., Eberman, B., Goddeau, D., Meyer, D. 2002.

Building VoiceXML Browsers with OpenVXI. In

proceedings the international WWW conference.

Kumar, A., Rajput, N., Chakraborty, D., Jindal, S.,

Nanavati, A. A., 2007. VOIGEN: A Technology for

Enabling Data Services in Developing Regions. IBM

Research Report No.RI0700.

Shao, Z., Capra, R., Perez-Quinones, M. A., 2003.

Annotations for HTML to VoiceXML Transcoding:

Producing Voice WebPages with Usability in Mind.

Kumar, A., Rajput, N., Chakraborty, D., Jindal, S.,

Nanavati, A. A., 2007. VOISERV: Creation and

Delivery of Converged Services through Voice for

Emerging Economies.

World Internet usage Statistics News and World

Population stats, viewed 17 September 2008,

<http://www.internetworldstats.com/stats.htm>

Voice Extensible Markup Language (VoiceXML) Version

2.0, viewed 3 September 2008

<http://www.w3.org/TR/voicexml20/>

Sri Lanka Internet Usage and Telecommunication reports,

viewed 8 October 2008

<http://www.internetworldstats.com/asia/lk.htm>

Asterisk: The Open Source PBX & Telephony platform,

viewed 13 November 2008

< http://www.asterisk.org/>

Ekiga.net :What are the service provided by Ekiga.net?,

viewed 26 November 2008, <http://www.ekiga.net>

Software Voip VoiceOne- Open source PBX on Asterisk.-

PBX software Voip, viewed 3 December 2008,

<http://www.voiceone.it/>

Lamb, M., Horowitz, B. Guidelines for a VoiceXML

Solution Using WebSphere Transcoding Publisher,

viewed 17 November 2008, <http://www-

3.ibm.com/software/webservers/transcoding/library.ht

ml>

FreeTTS 1.2- A speech synthesizer written entirely in the

java(TM) programming language, viewed 22

November 2008

< http://freetts.sourceforge.net/docs/index.php>

Moodle XML format- MoodleDocs, viewed 19 October

2008, <http://docs.moodle.org/en/Moodle_XML>

CSEDU 2009 - International Conference on Computer Supported Education

222