An analysis of screen reader use in India

1 1 2 3
Ted McCarthy , Joyojeet Pal , Edward Cutrell , Tanvi Marballi

1 2 3
University of Michigan Microsoft Research India New York University
School of Information Technology for Emerging Markets Polytechnic Institute
Ann Arbor, MI 48104 Bangalore, India 560 025 Brooklyn, NY 11201
{epmccart, joyojeet}@umich.edu cutrell@microsoft.com tanvimarballi@gmail.com

ABSTRACT 1. INTRODUCTION
We present the results of two surveys and a qualitative interview- Approximately 90% of the world‟s visually impaired live in the
based study with users of screen readers in India. Our early developing world1, and an estimated 15 million Indians are
interviews moved us in the direction of examining patterns that visually impaired.2 For people with vision impairments, access to
differentiate users of two particular software applications – the Assistive Technology (AT) can be decisive for participation in a
dominant market standard JAWS and the free, open source modern labor force where technology is increasingly ubiquitous.
challenger NVDA. A comparison between the two is timely and Since most AT for people with vision impairments tends to be
particularly relevant to issues elsewhere in the developing world. produced in the industrialized world, primarily for users from
In the short term, the question of choosing one application over those countries, there are problems of price point, support
another could be based on price and support for custom-made infrastructure, and language regionalization that can be hurdles
applications, but in the long term, issues of language support are for AT adoption in the developing world. As a result, screen
likely to be of concern as well. We explore software adoption readers that are free or low-cost and easily extensible to locally
behavior and present results that show the relationship between relevant software needs are of importance to the needs of the
the quality of audio and peoples‟ willingness to use one software developing world.
over another. We also compare the switch from JAWS to NVDA
to other kinds of switches from dominant software to open source Although some work has looked at the importance of low-cost AT
options. In conclusion, we discuss the business aspects of screen for the needs of the developing world [1], there has been little
readers and examine why the comparison between these two systematic investigation of the actual mechanics of low-cost
applications is particularly important in the discussion on options to dominant (and high-cost) software. Furthermore, there
accessible personal computing for people with vision impairments are few empirical studies that present data on the state of AT use
in the developing world. for people with vision impairments in any part of the developing
world. Despite the large community of persons with vision
impairments, their relatively small size as a „market‟ for AT
Categories and Subject Descriptors products has limited the amount of existing research on the
K.4.2 [Computers and Society]: Social Issues – Handicapped
technology use of this community.
persons/special needs.
In this paper, we explore the use of screen reader software by
General Terms people with vision impairments in India, and specifically examine
Design, Economics their behavior related to low-cost options on screen reading
technology. Screen reading software refers to programs that
enable blind or visually impaired users to better operate a
Keywords computer; they essentially replaces much of the graphical user
Disability, ICTD, Assistive Technology, Screen Readers, Open interface, allowing users to interact fully with a computer using
Source Software, JAWS, NVDA the mouse and receiving audio feedback (or tactile, if the
computer if equipped with a Braille output device – however,
none of our participants reported using a tactile output). The
screen reader employs a text-to-speech engine (TTS) that “reads”
out text from the screen, notifies users of the applications they are
Permission to make digital or hard copies of part or all of this using, their location within a spreadsheet or web page, etc. The
work for personal or classroom use is granted without fee quality of these TTS “voices” varies significantly, from very
provided that copies are not made or distributed for profit or human-like voices to voices that sound quite synthesized.
commercial advantage and that copies bear this notice and the full
citation on the first page. Copyrights for components of this work
owned by others than ACM must be honored. Abstracting with 2. APPROACH
credit is permitted. To copy otherwise, to republish, to post on We use a mixed-methods approach for the empirical data
servers or to redistribute to lists, requires prior specific permission presented here with input from a total of 200 respondents. In
and/or a fee.
1
ICTD '12, March 12 - 15 2012, Atlanta, GA, USA http://www.who.int/mediacentre/factsheets/fs282/en/
Copyright 2012 ACM 978-1-4503-1045-1/12/03…$10.00. 2
http://www.disabilityindia.org/djfactsoct07D.cfm
addition to an extensive review of the existing literature on screen Our efforts were primarily invested in sampling screen reader
reading technology, we studied forums of users and developers for users for two reasons – first, we see the open source survey as
screen readers, with specific attention to two particular software mainly playing a supporting role in explaining one of the various
programs – JAWS (Job Access with Speech) and NVDA (Non aspects relating to screen reader software preference in India, and
Visual Desktop Access) – which, between the two, have the secondly, the population of screen reader users in India is fairly
highest installation rates among people with vision impairments3. scattered (even in these two cities) and to sample a significant
enough population of assistive technology users was what we saw
Following our study of secondary data, we conducted in-depth
as the critical task ahead of us. There are a few biases in the
qualitative interviews of 20 users of screen readers, and those
populations sampled – for instance, sampling the open source
interviews were used to create two surveys; one specifically
software-using community through mailing lists dedicated to that
surveying screen reader use, and another on open source software
community likely biased us toward users who have a greater
use.
ideological motivation in their choice of technology. However,
The first survey examined a number of issues around screen because the survey compares application use across various kinds
reading technology use such as individual preference for one of open source software (operating systems, graphics applications,
screen reader over another, online activity, use of TTS, word processing applications etc.), we can control for this within
discussions of experience and typical problems with various the responses.
screen readers, and ability to complete specific tasks using screen
One important aspect of the survey was the distinction between
reading technology. The survey was conducted in-person in two
the locations of the interviewees based on the kind of survey
cities of India – Mumbai and Bangalore, where members of the
being conducted. In the screen reading survey, all the respondents
team were located - or online. Online respondents were reached
were based in India, although some of the experts recruited for the
through web-based forums on Inclusive Planet and Access India,
in-depth open-ended interviews were located outside of the
the first a social networking site and the second a mailing list,
country. For the open source software survey, we wanted to get a
both for persons with vision impairments in India. The survey was
sampling of non-Indian users as well as Indian users. The reason
fully accessible and could be completed using an online or
for this distinction, besides our need to focus on Indian users, was
desktop-based screen reading application – roughly a third of
that the process of procurement of and training in screen reading
respondents completed the survey online, and the remainder were
software differs in various parts of the world – so access to
surveyed in person. All surveys were conducted in English.
expensive screen reading software like JAWS (at about $1000 per
The second survey was very specifically aimed at sampling license) is mandated by various governments for both home and
behavior related to the switch from a dominant proprietary workplace use due to local disability-related laws. In contrast, use
software product to an open-source product. The decision to of most other software typically requires an investment by the user
conduct this survey came from early interviews in which we found (irrespective of piracy), and so the same problem was not
a sizable sampled population discussing an interest in moving applicable to the open source survey.
from JAWS, the dominant screen reading software, to NVDA, an
open source program and relative newcomer with a comparatively 2.1 Sample Description
smaller but rapidly growing market share. In this survey, our goal Both the open source software and the screen reading survey
was to understand economic versus ideological motivations in population were not random, therefore it is not clear to what
switching to open source software products generally to see what extent our sample reflects the rest of the population in that
comparisons could be drawn to screen reading software in category. For the screen reader survey, 34% of respondents were
particular. This survey was conducted via mailing lists of open female and 66% were male, while for the open source survey,
source software user mailing lists. 29% of respondents were female and 71% were male.
Ideally, we would have compared the results of the screen reading
survey with a pre-existing survey of open source product use, but
none such to our knowledge exists in India. Thus, the second
survey was performed with the purpose of fulfilling an important
gap in the literature.
There were 101 users sampled for the screen reading software
survey. The median age for respondents of this survey was 26, and
the average number of years of screen reader use was 5.4. 80% of
respondents own a screen reader for their computer, and 75% of
those surveyed use a screen reader on their mobile phones. About
half – 48% - of users who own screen readers reported that they
had obtained a pirated version, but it seems likely that this may be
a conservative estimate, though an attempt was made to ensure the
confidentiality of respondents‟ answers. There were 99 users
sampled for the open source software, 40% of whom were Indian
and 60% of whom lived elsewhere. Our sample was heavily male,
with that gender comprising 79% of the respondents.

Figure 1: Age distribution of Screen Reader survey

3
http://webaim.org/projects/screenreadersurvey3/ respondents (n=101)
Figure 2: Age distribution of open Source survey respondents Figure 4: Education level of open source survey respondents
(n=99)
In the case of open source survey respondents, this distribution
The age distribution of the respondents in both surveys skews does not matter significantly, though in the case of screen reader
relatively young, as can be seen in figures 1 and 2. One important users, the skew is indicative of the fact that the population is not a
difference between the screen reader survey and the open source good representation of vision-impaired persons in India more
survey is that the open source survey has a much higher generally.
proportion of younger, probably college-going or late school
respondents, whereas the screen reader-using population tended to In the results presented, we do not name any of the interviewees
be a few years past college, either late in their computer training who offered quotes, except those who are public figures and/or
or early in their careers. disability rights activists, and who agreed to allow their names to
be published.
If we look at the education levels of the respondents, we find
again that both populations are very highly educated, with a
majority possessing at least a bachelor‟s degree. 3. ANALYSIS OF RESULTS
The results from the surveys and interviews show that although
screen readers still form a fairly small market in India, there is a
fairly active community of users which largely show use patterns
comparable to some of the major global trends. We discuss here
some of the key findings from the surveys, and interpret them in
the context of qualitative research with assistive technology users
in Mumbai and Bangalore.

Figure 3: Education level of screen reader survey respondents

Figure 5: Level of expertise on screen reading software
We found that JAWS and NVDA were by far the most commonly
used applications – 85 of the total 101 respondents used JAWS
regularly, while 40 used NVDA regularly. WindowEyes was the
closest third at 6 users, making the battle for the Indian market
pretty much a two-way competition between these two
applications. However, we see that a proportionally higher country‟s intellectual property laws were at odds with the disabled
number of users were intermediate or expert at JAWS than they population‟s „Right to Read.‟
were at NVDA, indicating that many NVDA users tended to be
casual or experimental users.
“We have the right to read, but less than 5% of the
materials are available in accessible formats. What is
3.1 Predominance of pirated JAWS the right then? We will pirate the material; let us see
Despite its high cost, JAWS is practically ubiquitous among
who wants to sue this group.”
computer-using populations with vision impairments in India. In
the past it was common to come across people who used trial -Right to Read activist
versions of JAWS because of the unaffordable licensed version,
but due to the relatively high degree of software piracy in the
country, it is fairly trivial to obtain a copy of JAWS at an Software piracy was at roughly 64% in India in 20104, and Osorio
extremely low cost or completely free. has found that social beliefs around piracy can increase its
prevalence in a country [2]. The same research also finds that
illegal copying increases with high software price and lower GDP,
alongside which other research shows that software companies
operating in markets with a high degree of piracy have limited
incentive to build tools for those markets [3]. In short, there exists
a situation wherein JAWS is easily pirated, and the developer of
the software has no real incentive to try to reach the Indian
markets in the short term. The high prices of the software are in
part due to the structure of the industry, with few makers building
highly sophisticated software for a relatively small audience of
users, but more importantly, the software is frequently sold to
institutional buyers or governments in countries that have
disability laws. Thus the dependence on the end consumer‟s
ability to pay is limited.

JAWS has a history of dominance within the screen reading

market for more than the past decade, and the same applies within
India. The establishment of early adopters of screen readers in
Figure 6: Installed software (%) at place of primary use India quickly moved to JAWS, leading to most computer training
classes for screen reading to likewise teach JAWS. In the current
As we see in figure 6 above, over half the population stated that scenario, the switch away from JAWS has often been driven both
their primary screen reader was pirated. This of course does not by individuals motivated by an ideological preference for open
imply that the remaining 40% actually purchase their software, it source screen readers and by companies unwilling to use pirated
simply means it is a licensed copy – which could be a copy at a software for their vision impaired employees, which are then
place of work, a donated copy, or for that matter a licensed copy interested in finding workable alternatives.
of an open source software application (if they chose not to
classify that as free). When we asked users how many of them had
made a paid purchase of their screen reader, it was just 11% of the
sample, implying that the culture of acquiring paid versions of 3.2 Significant effects of ‘surface’ factors
screen reading software is not very prevalent. One of the most significant outcomes of the survey was the
relationship of TTS “voice” with the preference for a particular
software. We were surprised in initial interviews to find how
“Pirated JAWS is available so freely out here [in many people dropped NVDA very soon after their first attempt at
India]. Very easily. So, it's nothing to be proud of, but using it because of the audio “voice” quality of the TTS engine.
still, it's a fact of life that it's very freely available.
Hence people will not go into NVDA.”
-HM, Screen reader user, discussing piracy “The Eloquence [TTS] that comes default with [JAWS]
Mumbai is very simple to hear, people get used to it very fast.”
-Dr. Homiyar Mobedji,
Indeed, at roughly $1000 a license, just slightly below the 2010 Physiotherapist and screen reader user, Mumbai
estimate of the average Indian‟s per capita annual income, the
odds of a large number of people willing to pay for JAWS is on “If you’re going to be listening to [a mechanical voice]
the unlikely side. Of at least equal importance is the lack of a ten hours a day, it’s tough.”
significant social deterrent to pirating screen reading software.
This is evident in the approach of one important social movement Nirmita Narasimhan
of people with vision impairments in 2010, the „Right to Read‟ Lawyer and screen reader user, Bangalore
campaign in India. It took the normative position that since
Indians had the right to read, but the government nonetheless did 4
http://portal.bsa.org/globalpiracy2010/downloads/press/pr_india.
nothing to provide print-impaired people with this right, the pdf
 Figure 8: User preferences for screen reader features

We find strong statistical significance when looking at the figures

for preference for application support over voice quality as screen
reader users move from novice to advanced.
Much as we may find that the more one uses a screen reader, the
comparatively less one cares for the quality of voice, the initial
adoption factor makes an important impact. The importance of
voice quality for early stage users cannot be overstated. One
important factor reinforcing the preference for JAWS is cellular
phone users‟ familiarity with the output.

Talks uses Eloquence. And most people in India are

very familiar with Eloquence, hence Talks is very
readily accepted compared with MobileSpeak… if you
are given a choice of pirated softwares of Mobile Speak
and Talks, people would prefer Talks, in India.”

Figure 7: Common uses of screen reader software -HM, Screen reader user
Mumbai

And while respondents clearly marked their preference for screen Thus the fact that that Talks is the dominant cellular phone-based
readers with a natural voice (such as JAWS), the data about their screen reading application in India (used by 98% of our
preferences after becoming more experienced, and discussions respondents) further strengthens the preference for Eloquence (the
with advanced screen reader users, suggests something rather natural-voice speech synthesizer used by JAWS), and as users
different. The survey data shows that as users get more were unwilling to try out the cellular alternative MobileSpeak,
experienced with screen reader use, they speed up the audio they similarly dismissed NVDA because of the voice.
output to the point where it sounds very much like a mechanical
Up until recently, people have been trained using only JAWS or at
voice. In fact, some advanced users specifically stated preferring a
the very least „primarily‟ JAWS at training centers, and as 98% of
mechanical voice for the standardized intonation.
users surveyed cited JAWS as the one screen readers they first
As a user becomes an advanced screen reader user, the output used (one learned using NVDA and one using NVDA and
voice seems to matter less than the number of applications JAWS), they will likely continue to use this software well into
supported. The results of the survey show strongly (p = .010) that more advanced stages. Thus, even after one has ceased to value
advanced users find application support to be the most important the quality of voice, the loss in efficiency from switching to a new
quality in a screen reader, while novice users placed the most piece of software and re-learning a new interface often prevents
importance on the voice quality of the text-to-speech engine used this from occurring.
by the software.
Shown visually, the survey of screen reading suggests an inverse
“Since it's existed for quite some time, people are
relationship between level of expertise and valuation of features,
actually addicted with the Eloquence synthesizer,
as seen in figure 8.
because the quality of the speech is pretty good… So,
now if you ask them to transition from that speech
quality to a little bad speech quality [as in the case of
NVDA], it's actually the mindset issue, they cannot
really change it. It takes time for them to change. That's
the problem. They simply say that NVDA is bad just
because of the speech quality.”
Srinivasu Chakravarthula,
Yahoo! accessibility consultant, Bangalore

This can be seen as comparable to the behavior of people

switching from perhaps Windows to Linux or PC to Mac, wherein
there is a learning curve with which some users are willing to
experiment, but with the knowledge that there may be some
efficiency loss at least initially. More importantly, the lesson here
is that institutional factors for software adoption are probably
critical – so while a user may not ordinarily be willing to try out
an alternative to a certain piece of software, when thrown into an
institution which primarily supports such an alternative, the user
\ must adapt. Our survey shows that NVDA clearly follows JAWS
as the second-choice software and the software users are most factors. There are those who are willing to „experiment‟ with
likely to rate as being able to switch to if needed. Linux as they have more than one computer, and these users tend
to be more casual users, and another group that dedicatedly uses
3.3 Cost and software switching Linux either due to preference for the greater customizability, or
The range of factors influencing behavior relating to switching even on ideological grounds. Our results show that users are
from the dominant „industry standard‟ software of JAWS to lower significantly more likely to choose an open source application due
cost alternatives bear an interesting comparison to other to reasons related to cost if that application is not an operating
comparable switches. The results of the open source software use system, especially if it is an office suite, photo editor, or PDF
survey give us interesting insights into some of the influencing editor – that is, those applications which seemingly offer
factors. comparable services when found in open source when compared
with their commercial equivalents.

Table 1: Emphasis on cost as a factor for switching to an Open

Source technology The existence of a network of users is also fairly critical to the
adoption of any new software. In the case of screen reader use, we
Difference in found that the typical user frequently resorts to queries either from
Mean cost emphasis within one‟s immediate circle or through online forums. A lack of
Emphasis on from that a sufficient network of users has been discussed elsewhere as a
Application type Cost placed on OS common barrier for software switching behavior [4], including
Operating System 3.96 N/A specifically in the case of open source options to existing
dominant software [5-7].
All non-OS 3.39 p = .001***
These same network effects can be seen in our sample of mobile
Office Suite 3.14 p < .001*** phone screen reader users as well. In general, people were happy
Photo Editor 3.25 p < .01** with Talks, and chose it either because it was the first mobile TTS
they had heard of, or through networks of friends and colleagues.
PDF Editor 3.23 p < .01** For most people the choice of Talks, especially of that over
Audio Editor 3.54 p = .10 Mobile Speak, was often due to what was considered the market
standard because of referrals by friends. Several users explicitly
Statistical Software 3.75 p = .47 mentioned being comfortable with the voice quality of Talks, and
Video Editor 4.03 p = .80 said they were uninterested in switching.
*** = significant at p =.001; ** = significant at p=.01
For people with vision impairments, a screen reader functioning at
Table 1 shows the results of participant rating of cost as a factor in a sub-optimal level, even briefly, can have an extremely adverse
choosing open source rather than commercial software, rating impact on productivity. In India, given a relatively unfriendly
emphasis on a scale from 1 to 5. A score of 1 indicated that cost employment environment for people with disabilities, this is an
was a highly significant factor in choosing the open source even greater concern, because employees cannot take the risk of
product, and a 5 indicated that cost was of no matter in choosing being unproductive, albeit for a short period of time. The „cost‟
the open source software over a commercial approximate thus of the switch can be fairly significant.
alternative. The results show that there is much greater price
sensitivity for office applications, document management systems,
“I tried [NVDA] once very briefly, but the problem for
and graphical development applications compared with operating
me is if I have to switch to another screen reader,
systems in the decision to „switch.‟ Thus, the high cost of
effectively I need at least one week or ten days to
dominant software options in those categories - namely
familiarize myself with it and… for that period of time
Photoshop, Acrobat, or MS Office – are more likely to push a user
I'm not able to work productively because I'm still
towards a switch to a free version when compared with the switch
discovering these things”
between a commercial and open source operating system,
especially when that version offers the most commonly used NM, JAWS user, Bangalore
functionalities. In contrast, we found that proprietary specialized
software, such as SPSS statistical software, is less driven by price.
Several of our interviewees explicitly mentioned the importance
We explain this as a factor of user expertise – thus document of the large network of extant JAWS users, the extensive software
management or office application software can be qualified as documentation available for the product, and the general
generic, therefore used by a fairly wide number of users. In entrenchment of the software within the visually impaired
comparison, video editing software or statistical packages are not community, all of which exist to a much lesser degree in the case
likely to be evaluated easily in a framework such as ours, because of NVDA. The switch to NVDA (there were very few instances of
we find far fewer „expert‟ users in a random sampling. That is to any other open source software discussed by the users, so we
say, those users whose professions depend on SPSS or video restrict our analysis to this comparison) came with the fear of not
editing software, and only those users, are likely to have much finding the right support environment, especially if the work was
lower price sensitivity or impetus to switch to open source mission-critical.
options, especially if their professions depend on it.
What is perhaps most interesting is that we found the Windows to
Linux switch to be significantly less driven by cost than by other
 “Perhaps the largest barrier to NVDA use is that
eSpeak [the default TTS for NVDA] is not being
introduced at the training center level”
Dipendra Manocha, Screen reader user
and disability rights activist, Mumbai

In our own sample as well, JAWS was clearly the dominant

preference (100% of the screen reading sample, in fact). While the
NVDA users were far outnumbered by the JAWS users, the
NVDA users were also slightly more technical. More than a third
of the high-frequency NVDA users (5 out of 14) wrote scripts for
their screen readers, whereas just about a fifth (6 out of 29) of the
high frequency JAWS users from our sample were writing scripts.
The results are not statistically significant, but nonetheless
underline the notion that feeds the idea that one needs to be more
technical to be an NVDA user.

Figure 9: Top ranked sources of information on screen readers 3.5 Screen reading in the work environment
A factor not immediately obvious on switching behavior is that
social networks play an incredibly important role in vision “Middle-level and startup companies… cannot afford
impaired computer users‟ technology choices. This emerged to spend about 1000 dollars [per license of JAWS] … If
somewhat in our interviews with people, but as we see with the people start using open source screen readers like
top sources of information on technology in figure 9 above, online NVDA, it's easy to convince an employer to give an
sources (typically social networking forums), organizations employment opportunity to these candidates, and it's
(typically NGOs) and others (typically friends) are top sources on easy for them to get in onto the payrolls… I recommend
issues relating to screen readers and AT. In other words, new our trainees to learn using NVDA because it's easy for
technologies spread very quickly through word of mouth sources us to generate employment for them.”
and social networks.
Srinivasu Chakravarthula,
Yahoo! accessibility consultant, Bangalore
3.4 The discourse of complexity
One of the motivations behind exploring the parallels between By 2010, Indian software companies had started investing in
JAWS/NVDA and perhaps the most easily comparable condition developing NVDA for compatibility with their internal systems.
of Windows/Linux was the perception of complexity. As opposed This is an important direction for the future of screen reading as it
to the Windows OS packaged with most off-the-shelf computers, indicates a recognition among companies that licenses for
there is typically an additional effort involved in switching to expensive screen reading software may not be viable in the long-
Linux. The perception that the average Linux user needs to be run. With the implementation of the United Nations Convention
more technical is fairly pervasive, as the Linux OS is generally on the Rights of Persons with Disabilities and India‟s own
perceived as having a steep learning curve, associated with an disability-related legislation, it is clear that larger corporations are
early group of command line hackers [8]. This made it such that likely to see an increase in employees with vision impairments.
when user-friendly releases of Linux (e.g. Ubuntu) arose, there Alongside this, the increasing implementation of intellectual
was a need to market these as usable by non-geeks [9]. property laws in the organized sector has meant that „Pirated
This association with the hacker ethic has also contributed to the JAWS‟ is not a serious option for the workplace, even if
idea of open source software use as itself being tied to the intent employers choose to turn a blind eye to whatever is installed on
of, or active participation in, the further development of their employees‟ personal machines. Thus stable versions of
applications [10] or at the very least in user groups [11]. NVDA that can handle custom-made internal applications are
Arguably, this „geekification‟ plays a role in the idea that any likely to grow in prominence among major employers.
open source software requires a greater ability to deal with
complexity and the minutiae of software functionality.
“People aren’t seeing the large picture – eventually
To some extent, this was reflected in computer courses for they’re going to have to pay for [pirated versions of
visually impaired populations focusing only on JAWS, much in JAWS] – if they get a job, or… whether out of their own
the way that introductory computer courses do not stray beyond pocket, or through a company, or something else.”
Windows. The few computer courses that do offer training in
NVDA (Enable India in Bangalore, for instance) offer it as Dipendra Manocha, Screen reader user
secondary to JAWS, usually spending very little time on it. and disability rights activist, Mumbai

3.6 Regional Language and Accent Support

Though there are many millions of speakers of the 22 official
languages of India, this distribution of speakers is not
proportionately represented in the digital realm [12]. Though this
may be due in part to the fact that wealthier Indians tend to have a want it in Indian language… there's a demand.
higher command of English, further research is necessary to Growing demand, really really growing demand for this
definitively conclude to what extent this is the case. Nevertheless, [Indian language screen readers].”
the development of TTS engines in lesser-spoken languages has SN, screen reader trainer, Bangalore
piqued the interest of the research community [13] [14] as well as
a great number of organizations, activists and individuals in the
vision impairment community in India. Indeed, “high-quality” TTS for local languages brings up an
important distinction: it is not enough to merely build a TTS that
works for smaller languages, but this TTS must be usable, or else
“I'd be quite happy to switch to another screen reader people will fall back on using one of the more established TTS,
with an Indian language TTS once that gets up and even if it is not developed in their preferred language. Tucker and
working” Shalonova show that, while the implementation of a functional
NM, screen reader user, Bangalore local-language TTS can be easily undertaken, particularly with
open source TTS such as eSpeak and Festival, it takes significant
time and expertise to produce a local-language TTS of high
Within India itself, a great number of NGOs and government enough quality such that it is likely to be successfully used [13].
organizations are working toward the development of TTS for
smaller languages. In the past, the DAISY Forum India has Findings from our survey support this: in fact, all 9 of the
partnered with local NGOs to develop local-language support, and languages or accents participants said they would prefer to use
an agreement was recently made to do the same for the over their current TTS already exist. Of the 22 respondents who
development of several more languages, including Gujarati and had used a TTS in their preferred language, but did not regularly
Marathi. Additionally, the TTS Consortium, a research group do so, 14 (64%) responded that they did not use this TTS because
consisting of several IITs (Indian Institute of Technology) from of its insufficient quality, whether due to poor “voice” quality or
different cities as well as C-DAC (Centre for Development and because of insufficient synthesizer speed or performance.
Advanced Computing), a government-funded research group, is Additionally, of those 13 who provided a reason as to why they
currently involved in a project to develop a number of regional- had not tried a TTS with the language of their preference, 11
language TTS.5 respondents stated that they had been unable to gain access to this
TTS, often stating (incorrectly) that TTS for this language does
While text-to-speech engines exist for a number of Indian not exist, that it was because they did not know where to find the
languages, these are often of poor quality or too expensive for software, or even because they were unable to figure out how to
most users. Preference for Indian English accent TTS among install the TTS or make the proper settings change to enable their
Indians (and particularly that matching the subjects‟ local accent) screen readers to utilize the TTS. While this is a very small
is higher when compared with a US English accent TTS, the sample (N = 13), the fact that 11 of the 13 users who stated that
accent most readily available, and therefore most commonly used, they had not used a TTS language of their preference listed a
in screen readers in India. Additionally, findings show a decrease reason other than not wanting to use it (even when most of those
in intelligibility for those users with worse English when using the respondents knew that the TTS existed) shows that there is much
US English accent TTS [14]. This points to an especial need for to be done to increase access to regional language TTS besides
high-quality local language or Indian English-accent TTS for simple development. Users must be made aware of the existence
those with poorer English, which likely includes many members of these TTS, and must be able to access and easily use them as
of the visually impaired community in India, given the well.
correspondence between that population and higher rates of
poverty in some parts of the developing world. Taking all these findings into consideration, it appears that the
mere development of a “functional” TTS for small languages is
The results of our own survey also showed a desire for regional not enough; these TTS must be of high quality, well-marketed,
language TTS development in India - 52% of respondents and easily accessible, both from the standpoint of obtaining the
indicated that there is some language or accent in which they software and from that of actually using it.
would rather use their screen readers, and these answers were
spread among 8 different Indian languages, as well as the Indian
English accent. With the realization that 98% of respondents “There are no standards for Indian languages… there
indicated that they primarily use a screen reader with either is no standard keyboard layout.”
American or British English, and only one user regularly utilizes a SM, former SAFA developer and screen reader
screen reader with an Indian English accent, it seems that there is consultant, Bangalore
indeed a desire within the visually impaired community in India
for high-quality TTS in local languages.
While there is evident desire among the visually impaired
community in India for local language TTS and screen reading
“So, in the first act, people are forced to use screen software, there are a number of obstacles presented by this
reader English, for English screen reader. And when it challenge that are likely not immediately apparent, even once the
comes to a specific reading purpose, there is where they difficulties presented by the development of high-quality TTS are
addressed. For instance, the lack of standardization in Indian
computer hardware came up as a problem in several interviews
5
http://www.cdacmumbai.in/index.php/research_and_publications with developers who had been involved in the SAFA (Screen
/projects/text_to_speech_synthesis_systems_for_indian_langua Access For All) project to develop an Indian-language screen
ges_tts_il reader. Among other complaints, one that was mentioned
numerous times was the number of computer keyboard layouts support for their internal applications on NVDA. Online forums
used for Hindi alone: developers interviewed described the show that there are regular updates of new applications being
number of Hindi keyboard layouts as lying somewhere between supported or minor adjustments to improve the product occurring
15 and 21 different configurations, all of which would require on a regular basis.
individual attention to be fully supported by the screen reader One of the features that made NVDA particularly attractive to
software. (The SAFA project settled on developing support for users was that it can be used with a USB stick without the need
three of these.) for any further software.
Additionally, the concern over the general lack of Indian language
screen reader-friendly fonts (namely Unicode) was brought up
often. Though most screen readers are written to support Unicode, “With JAWS, because it has to load its video training
many websites, including popular Tamil and Hindi news sites manager, you can't load it through a USB stick… The
such as Eenadu and Jagran, use ASCII fonts, and therefore portable version of NVDA, especially for people who
remain inaccessible, even to most screen readers that support are going around in cybercafés and using computers of
these languages [15]. their friends, or on the university campus, or anywhere
like that, it's definitely a big advantage that you can
easily just stick in your thumb drive and start using
“Most TTS’ support Unicode based fonts. However, NVDA on the move.”
fonts in Indian languages are older than Unicode, and Dr. Homiyar Mobedji,
many of the popular publishing industry fonts are still Mumbai
not following Unicode coding system.”
DM, former SAFA developer, Bangalore Given that the vast majority of vision impaired computer users in
India do not have access to their own machines, this is quite
Though a report from Google in January 20106 showed Unicode important, though not something one foresees competitors
support across the web nearing 50% and growing quickly, this ignoring for long. It is, however, the apparent momentum
alone represents a challenge to local language TTS that may be surrounding NVDA that may be its strongest advantage.
easily overlooked. Even with the development of high quality, NVDA‟s funding model has an “OSS 2.0” flavor to it [16]. That
easily accessible regional language TTS engines and screen is, while the older OSS model was personified by Linux, Perl,
readers, these tools are only as good as the material they are trying etc., each of which had many contributors and little explicit
to access. organization, the newer OSS 2.0 model has few developers
receiving feedback from a larger community or users, often
4. Discussion receiving funding from outside sources. NVDA does exactly this
Our initial interviews that would lead us to creating the two – though OSS 2.0 is also described largely as being for-profit with
surveys we deployed had already indicated that the JAWS vs. value-added services such as Red Hat or IBM‟s Star Office
NVDA question was likely to be central to any discussion of the (compared with totally free OpenOffice.org).
future of screen readers in India. The importance of this question
Besides the general support from the online community, there
is only reinforced by the findings from the survey.
have been a number of proactive steps that have pushed the
JAWS remains by far the dominant screen reader in the world, development of NVDA further. First, a significant number of the
holding over 67% of the market share.7 In our own sample, almost development team for SAFA (Screen Access For All), an Indian
all the interviewed expert users had JAWS as their primary screen language screen reader, were reassigned to work on NVDA
reader, even when they considered themselves advanced NVDA (according to personal communication with Dipendra Manocha,
users. Yet, it is not clear that this dominance will remain the case coordinator for the DAISY Consortium). The Saksham Trust, a
perpetually. NVDA has only existed for a fraction of the time New Delhi-based NGO, recently gave away 200 netbooks with
JAWS has, and is gaining ground quite rapidly. The WebAIM NVDA pre-installed to visually impaired persons “who want to
surveys show that there has been a growing move towards open use the computers for their daily reading and writing purpose, but
source software use among screen readers around the world, with due to the high cost of hardware and software, they are unable to
NVDA the fastest gaining application in the screen reading do so.”9
category, currently outranked only by JAWS in terms of the
Finally, the possibility of local language development has the
number of installed machines worldwide.
potential to be a very important factor as time goes by. While the
As we have pointed out here, there are both business-related and reality of the day is that computer users in India can manage with
technical reasons for why open source software may make a very (and probably prefer) English language interfaces, this is likely to
strong impact starting in the developing world. NVDA‟s website change in time, and the day we see screen readers with high
lists the Mozilla Foundation, Adobe, Yahoo!, and Microsoft as quality Hindi, Telugu, or Bangla output may not be far.
significant donors to the project8, in addition to which there have
For commercial screen readers like JAWS, the situation is not
been a number of firms encouraging NVDA use in place of
necessarily ominous; these findings may only mean that business
pirated software on their machines, and in doing so, building
models for these markets be different. As in the case of other
major software firms, hard questions on pricing strategies for
6
http://googleblog.blogspot.com/2010/01/unicode-nearing-50-of- international markets will have to be dealt with, and in an
web.html increasingly competitive market that corresponds to the ability to
7
http://webaim.org/projects/screenreadersurvey3/
8 9
http://www.nvaccess.org/blog http://www.saksham.org/project.php?id=11
support custom applications. This may require that commercial 6. REFERENCES
screen readers expand their developer networks.
Finally, our study on screen readers also points to directions [1] Pal, J., U. Vallauri, and V. Tsaran. Low-cost assistive
ahead for major software and web service companies or operating technology in the developing world: a research agenda for
systems. For instance, one of the most frequent complaints of information schools. 2011. ACM.
NVDA users was the lack of appropriate support for office
applications. Interestingly, Microsoft, with its own OS-based [2] Osorio, C.A., A Contribution to the Understanding of Illegal
screen reader, Narrator, may also be an important player in this Copying of Software. Levine's Working Paper Archive,
space over time (only four respondents regularly used the 2003.
Windows-based Narrator, the same number that used the Linux- [3] Maskus, K.E., Intellectual property challenges for
based Orca) developing countries: An economic perspective. University
. of Illinois Law Review, 2001: p. 457.
“NVDA and JAWS comparatively, JAWS will work with [4] Katz, M.L. and C. Shapiro, Technology adoption in the
many more applications... especially the formal ones presence of network externalities. The journal of political
like Word, with Microsoft Office.” economy, 1986: p. 822-841.
Dr. Homiyar Mobedji, [5] Huysmans, P., K. Ven, and J. Verelst, Reasons for the non-
Mumbai adoption of OpenOffice. org in a data-intensive public
administration. First Monday, 2008. 13(10).

5. Conclusion [6] Kelley, J., R. Patnayakuni, and Z. Xiang, " OOo! I want
that!" An Examination of Individual Adoption of OpenOffice
Assistive technology use related to vision impairment in India (or
Software. AMCIS 2006 Proceedings, 2006: p. 109.
in the developing world broadly) has had very limited serious
scholarly interest until recently. Our purpose in this study was to [7] Abdelwahab, M.M.A., Assessing the switching barriers
highlight some of the important issues around screen reader use in between Microsoft Office and OpenOffice. org, 2009,
India in the hope that in time the subject becomes an important Carleton University.
enough area of concern within the rapidly growing ICTD [8] Himanen, P., L. Torvalds, and M. Castells, The hacker
community. ethic2002: Random House Trade Paperbacks.
A number of the findings here give us an insight into how screen [9] Grant, R., Ubuntu Linux for non-geeks: a pain-free, project-
readers are being adopted in a country where the proliferation of based, get-things-done guidebook2006: No Starch Press, San
assistive technology within the population with vision Francisco.
impairments is still low, and in time as research in this space
expands, we will be able to build on some of the findings [10] Lakhani, K. and R. Wolf, Why hackers do what they do:
highlighted here. Understanding motivation and effort in free/open source
software projects. MIT Sloan Working Paper No. 4425-03,
For one, the issues on piracy that we highlight are likely to have 2003.
deep impacts on the market for screen readers, and as we discuss
in section 4, potentially spur an investment into NVDA or other [11] Bagozzi, R.P. and U.M. Dholakia, Open source software user
open source screen readers in accordance with the OSS 2.0 model communities: A study of participation in Linux user groups.
of development. Management Science, 2006. 52(7): p. 1099.
The two instances in our study where we find the most significant [12] Pal, J., The developmental promise of information and
findings also hold potential for follow-up work. The findings communications technology in India. Contemporary South
around cost of software switching from dominant software to open Asia, 2003. 12(1): p. 103-119.
source software give us an area of further investigation that can be [13] Tucker, R. and K. Shalonova. Supporting the creation of
applied not just to screen reading, but more broadly to a range of TTS for local language voice information systems. 2005.
other software applications. Finally, our findings on the surface
[14] Weber, F. and K. Bali. Real Voice and TTS Accent Effects
factors relating to screen reader preference need to be validated
on Intelligibility and Comprehension for Indian Speakers of
with screen reader use patterns in other parts of the world.
English as a Second Language. 2009.
Through this, we hope the conclusions of this work are relevant
not just to assistive technology in the developing world, but to the [15] Raghavendra, E.V. and K. Prahallad. A multilingual screen
wider community of AT users across the globe. reader in Indian languages. 2010. IEEE.
[16] Fitzgerald, B., The transformation of open source software.
Mis Quarterly, 2006. 30(3): p. 587-598.