Press "Enter" to skip to content

Swarachakra – A Zero-Budget Touch Keyboard for Indian Languages | Anirudha Joshi | TEDxDharavi


[Music]
I’m going to begin by asking you a
question how many of you have typed
something since this morning okay almost
everyone has put up their hand and how
many of you have typed something since
this morning in your mother tongue yeah
two three four people have put up their
hands so this is a question that I have
asked in many or to many audiences in
many countries and only in India do I
get as such a poor response the second
question that I ask so that is what I’m
going to talk about today so we come
across this problem from time to time
everywhere so for example once somebody
had put up this beautiful mural or in
IIT Bombay campus it is very delicate
and very nicely done and because it is
so delicate he put up a sign next to it
and the sign said please do not touch
the English sign was beautifully typeset
the Hindi sign was handwritten we come
across this kind of a situation very
often in our lives it is not only about
these homemade signs it is also about
professionally designed film posters so
these are professional designers but
they do not know how to type in Hindi so
they tend to put up a Hindi poster also
in the Roman script and sometimes there
is a little bit of Hindi but it is very
small and on one side so typographers
really don’t play that much with Hindi
typefaces as much as they do with
English typefaces and this is a problem
that actually bothers me quite a bit
this is on the Wikipedia so though this
graph shows that along the x-axis it
plots the languages according to the
number of speakers in millions and along
the y-axis it shows the number of Viki
parity Wikipedia articles in thousands
okay
in this corner here are languages like
English German Russian Spanish
Portuguese Mandarin these are all
languages with very high number of very
large number of native speakers as well
as a very large number of Wikipedia
articles both about or about 100 million
native speakers and about above a
million wikipedia articles this corner
on the left we have languages which I
have a very small number of native
speakers and still a very large number
of Wikipedia articles probably have
never heard of this language called
worry worried in India it is the
language spoken in Philippines it has
only about 2.6 million speakers only 26
lakh speakers so that’s between under a
as alpha and got Cooper so that much
population but they have more than a
million speakers so that these are
beautiful languages but many of our
Indian languages unfortunately are sort
of here in the middle okay so the only
three Indian languages with more than
one lacks Wikipedia articles 100,000
Wikipedia articles so who is the highest
with 130,000 Hindi is the second-highest
with 123 thousand and Tamil is a third
with hundred and fifteen thousand and
all other languages Malayalam Marathi
Kannada Gujarati Punjabi Bangla are less
than hundred thousand in fact Oda is
down here with only 2,500 articles in
the Wikipedia so that is something that
bothers me
so people have asked why is it that we
are working in this space we could just
teach everyone how to type in English we
have our schools we commonly teach
English in our schools so this should
not be a problem in a few years well the
problem is that technology penetrates
society much faster than we can teach
people okay so for example this red line
here shows the number of mobile phones
back from 1991 till present in India in
crores so we have about 118 crore mobile
connections in India right now okay and
that is our population so most of our
population is almost covered okay this
yellow line here which you can probably
dimly see shows the number of Internet
users in India
and that is rising very rapidly also and
this blue line after that shows the
number of smartphone connections in
India smartphone devices being sold in
India and this green line shows the
number of graduates in India okay
HSC diploma metric the pink line there
is the metric middle school primary
school and that is the literacy line and
we can see that technology new
technology particularly penetrates
through the layers of society very very
quickly and there is a large chunk of
what we call as non-english but literate
population these are extremely
productive members of society but they
cannot read in English and they are that
those yellow and the blue lines are
almost touching them they are already
penetrating in that circle and this is
the current and now problem they can
read in their own mother tongues so
teaching them English is not going to be
a viable option in the next few the
second question that sometimes people
ask me is okay we cannot teach people
English but we can use the ideas from
these languages into our own languages
so this world map sort of shows which
what are the different types of scripts
that are used in the world
so this grey part is where what are
called as alphabets these are all so the
Roman script that we use in English is
the alphabet so this grey part Europe
America’s southern half of Africa and
Australia and many other countries use
the alphabets this red part on the right
is China Japan Korea so those are the
ones which used logo graphic and feature
languages this middle part over here are
the languages such as Persian Arabic our
own or do so those are the absurd
languages and the languages from which
we can borrow from are spoken in these
countries so India Nepal Sri Lanka
Bangladesh Burma and a few countries
around us Ethiopia and some parts of
northern Canada the other North Pole so
most of the people who speak these
languages are in these countries and so
if any new ideas have to come they have
to come from
here we cannot really borrow from very
far away so given that this is the
problem that people cannot type what is
it that these people cannot do they
cannot make presentations like the one
that I am making they cannot write
documents they cannot create posters or
shop signs they cannot send a message
they cannot shop online or they cannot
book a train ticket
or search or even save a contact so we
are preventing people from doing all
these things just because we are not
able to give them a usable keyboard so
this is the design problem that we have
been working on for all these years
given that typing in indian languages is
hard how to make a keyboard that is
usable by every literate indian so that
was our dream that was our goal so we’ve
been working on this problem for quite a
few years we started back in 2002 by
designing a new keyboard a completely
new design for the desktop keyboard we
continued when smart for the band
feature phones started arriving so 2008
2009 we were working on feature phones
but today I am going to talk about our
work on smartphones so around 2010 smart
phones started arriving and that is when
this bright young man came to me and
said that he wanted to design a keyboard
for smartphone so that’s where it
started he came up with several designs
but one specific design he came up with
is swara chakra and I like I’ll talk
about this design here so many people in
IIT and outside liked his design that
soon we had a large number of
contributors so we currently have about
90 contributors for contributing to
development maintenance and design of
Sri chakra in various languages so I
will just try to explain how we went
about it
so there I essentially looked at it as
four important problems the number of
characters the sequence of typing the
conjunct and the speed with which we
type so just four important problems to
solve a design so the first problem is
the number we just a large number of
characters larger than the Roman scripts
anyways so the Roman script for example
has 26 characters most Indian languages
would have between 55 to 65 characters
too
so that makes it harder how did we
tackle it
well that continues to be hard but we
tried our best we use the familiar
layout so this is the layout that you
use in kindergarten to teach children
how to how to read we use that layout
the group characters so that they are
easy to find we put some you use our
best graphic design techniques to make
the layout beautiful the second problem
that people face is there are sequence
how do I invert sequence so if I will
type car were type the curve first or
office now that is simple it looks like
the matra will come second the consonant
will come first but what if I had to
take the key
does the mattre come first because the
matter appears first visually in fact
our mattress they are appear on all
sides of the consonant to the left to
the right above below so this is what
confuses a lot of people so what we try
to do is we try to take the decision of
sequence out of the hands of the user
what we do is you should touch the
keyboard you should touch a particular
consonant on your touch the keyboard
displays the circle of all the vowels
that is why it is called the soirée
chakra it’s the chakra of vowels so you
take the decision of sequence out and
you just select the one that you want at
that time the third problem is of
conjunct so for example if you are
typing clumsy in English you will just
type clumsy right and the C and the L
will stand next to each other without
touching we are a little more friendly
in India so the co and the law will join
together and will form a new character
and this is what confuses a lot of
people and kirai kirai and love is
joined together but they will still look
like clue i mean current law you can
read in them but in some cases like when
you are doing koh-i-noor role for
example the row disappears or when
you’re typing recur root the raca the
ref our people think they should type
after with a type the core like all
matras but actually have to type it
before so these are the difficulties
that a lot of people faced so what we
did was
we made this process a lot easier so for
example if you want to type a conjunct
starting with BRR you touch BRR and then
you move your finger up to add of the
Halong as soon as you do that the
keyboard shows you all possible conjunct
that you can make
starting with bow and then you have to
just choose so if you trying to type
blah you just search for blah blah and
then you go in that direction and you
get blah so it makes typing a conjunct
lot easier we need a few other tricks
which I will not go into because I don’t
have that much time but we put all these
design ideas together and we released so
a checker for various languages hindi
marathi and bengali gujarati Malayalam
Kannada Telugu and Odia oh so 12
languages is what we are supporting now
all of these languages are being
supported by various groups through
voluntary efforts and what is we
launched the product in 2013 and in the
last four and a half years it has
reached 3.1 million people 31 lakh
downloads in the last four and half
years now one thing I must mention about
this keyboard it is it is a zero budget
keyboard it runs on voluntary
contribution of people’s effort not
money so for a we don’t advertise we
don’t have budget for advertising so the
keyboard spreads only through the
word-of-mouth and so we are we had
extremely we were extremely surprised in
fact when we got our first 50,000
downloads in the first year we were
thinking who the hell are this is a lab
based product we just put it out so that
it is easy to download why are people
downloading it and are they even using
it so we check that for the first time
after one year and we found that winning
one year just the group of fifty
thousand people had typed about 13 lakh
words okay over the next couple of years
that rate increased it went from 13 lakh
words in one year to 20 crore words in
three years so the usage was like
blowing up so earlier every user was
typing about 10 words per month using
our keyboard in a couple of years they
started typing 100 words a month using
our keyboard so the first three we are
quite satisfied we have solved the
problem that people can now type to
solve the last problem which is of speed
what we did is we set up a race between
four keyboards that were available and
that were popular at that time so these
were soiree chakra was one of them we
took in c-dac in script which is the
government standard keyboard implemented
on Android and then we took fifth key
which is a top-of-the-line keyboard with
lot of latest technology in it and we
did Spurs I must point out that both
Swift key and Spurs also had things like
prediction and see if she also had
swiping and so on which are some of the
latest technologies so we set up a race
of these four keyboards and Miss what we
decided is that we will do this race
which
children so children helped us identify
which would be the best keyboard so we
went to government schools for
government schools too in Mumbai and two
in one in London and one about 15
kilometers out of nanda in the village
so we went to different kinds of schools
because we wanted different kinds of
representations and we trained the
children each child we trained
individually for about 40 minutes and
then we did a first-time usability test
with them to make sure that they could
they had actually learned what they had
taught them and then after that we
called them for about 15 days every day
for two sessions for 30 sessions in all
in each session they came and typed out
ten phrases and then we counted how fast
are they typing and how well are they
doing and so this was the result people
started typing around 15 characters per
minute now that’s very low typically you
type for example at about 100 characters
per minute in English but within about
four or five hours they’d reached about
4045 characters across the different
keyboards one thing I would say is that
swear checker came out as the winner
because that yellow line at the top that
you see that is the sriracha Caroline
and so laughs it was consistently ahead
of the other keyboards so it won the
race and we were happy about it but
there was one thing that we were not
very happy about one of the things that
we wanted to work on next was to add
prediction to swat a checker because we
saw all these other people using
fantastic prediction systems and what we
were surprised to find is that both the
keyboards that were using prediction in
the latest staff were actually slower
than the other ones keep odds did not
predict about half the word so the gray
stuff about were the words that were not
predicted and people did not select even
the words that were predicted so these
dashed lines that you see in between for
the words that they did not select so
period people selected a very small set
of words did they will did they do well
in the words known so people actually
the same person typing without using the
prediction was typing faster and the
person who was using the prediction so
this brings me to our current work that
we are doing right now so in the last
few years speech in Indian languages
become
Jabalpur Johnny key ticket get nakey
jangan man a denial Jaya Hey
so speech is good it is improving
dramatically but it sometimes does not
capture everything that you are saying
correctly so you have to go and make
changes and some time and it is shown
that in English that slows people down
as much that it is not so effective so
we need a similar step and what we found
was that as soon as we gave people
speech their spirit went up so the red
line shows one group and which was
piping at around 40 to 45 characters per
minute the blue line shows the other
group and when he gave the other group
again they also end up and this group
came down so speech helps in improving
the speed so that is what I wanted to
leave you with one is that a large
number of Indians have just started to
type they found a new voice second
prediction is not necessarily working
for Indian languages at least third
speech seems to help a lot and I believe
that we have a huge opportunity coming
up to design for new group of users that
has found a new voice and the lastly I
would invite all of you to contribute to
swear chakra and help it grow thank you
very much
[Applause] [Music]
Please follow and like us: