0:00:08.334,0:00:09.417
Hi.

0:00:10.562,0:00:13.951
We are living in an exciting era,

0:00:13.952,0:00:19.281
where innovation and technology[br]has the potential to do the unimaginable,

0:00:19.282,0:00:22.559
and it becomes even more unimaginable

0:00:22.560,0:00:26.560
when it breaks down the gaps[br]between disability and ability.

0:00:28.345,0:00:31.325
15% of the world population

0:00:32.564,0:00:35.324
- 1 billion people around the world -

0:00:35.325,0:00:37.184
lives with disabilities

0:00:37.185,0:00:41.668
which makes people with disabilities[br]the largest minority in the world.

0:00:42.605,0:00:45.264
And they are not living[br]on a different planet.

0:00:45.265,0:00:50.145
They may be part of our families,[br]friends, or colleagues.

0:00:51.426,0:00:55.985
Today, I'm going to tell you[br]how people with speech disabilities

0:00:55.986,0:00:59.366
will have a way to better communicate.

0:00:59.375,0:01:03.233
I was 7 years old[br]when my sister Amal was born.

0:01:03.234,0:01:05.893
I was too young to see the challenges

0:01:05.894,0:01:09.463
that my family was facing[br]on a daily basis,

0:01:09.464,0:01:13.813
but I could see that Amal[br]couldn't crawl, or eat, or talk

0:01:13.814,0:01:16.913
like any other baby her age.

0:01:16.914,0:01:22.063
But with time, we adjusted[br]to raise a baby with cerebral palsy,

0:01:22.064,0:01:26.492
while understanding her special[br]communication patterns and needs.

0:01:28.406,0:01:29.845
Nine years later,

0:01:29.846,0:01:33.469
my family was blessed[br]to have another baby, Ahmad.

0:01:34.469,0:01:38.288
Ahmad decided to grow up[br]exactly like his sister Amal,

0:01:38.289,0:01:42.838
being so smart, so sharp,[br]curious about everything around him,

0:01:42.839,0:01:47.208
but he also decided to invent[br]his special communication patterns

0:01:47.209,0:01:48.809
to communicate with us,

0:01:49.782,0:01:53.081
and for the other people[br]that couldn't understand him,

0:01:53.082,0:01:55.208
we had to translate.

0:01:55.209,0:01:59.626
Amal and Ahmad say "num"[br]when they are hungry,

0:01:59.659,0:02:04.528
and they say "ahh" to call[br]the name of Nora, my sister.

0:02:04.542,0:02:08.833
And when they want to call[br]my name, they say "abeya".

0:02:08.834,0:02:12.585
In case they want to go[br]to the bathroom, they say "kkhh".

0:02:13.366,0:02:16.945
We understand most[br]of their special communication patterns,

0:02:16.946,0:02:20.546
but it's only us, the close circle.

0:02:20.551,0:02:25.131
And this is the case for most[br]of the people who have an unclear voice.

0:02:26.292,0:02:29.471
One of those people is Urit.

0:02:29.472,0:02:33.691
Urit is a 34-year-old woman[br]with cerebral palsy.

0:02:33.692,0:02:35.946
She is living an independent life.

0:02:35.947,0:02:41.003
She can drive her car, go to the gym,[br]and do a lot of other things.

0:02:42.917,0:02:47.656
However, when it comes[br]to communicating using her voice,

0:02:47.657,0:02:50.912
sometimes, it can become[br]harder than going to the gym,

0:02:50.913,0:02:53.122
and more frustrating

0:02:53.123,0:02:58.542
because she finds herself repeating[br]the same words again and again

0:02:58.543,0:03:01.067
in order to be understood.

0:03:01.068,0:03:04.738
We asked Urit to say[br]a few words in English.

0:03:06.370,0:03:08.199
Let's listen to her together

0:03:08.200,0:03:10.790
and see if you can understand[br]what she's trying to say.

0:03:11.856,0:03:13.946
(unclear speech)

0:03:17.481,0:03:21.861
I don't know how many of you[br]could understand her this first time,

0:03:21.862,0:03:23.471
but let's listen to her again,

0:03:23.472,0:03:27.521
and really focus and try to understand[br]what she's trying to say.

0:03:27.522,0:03:29.488
(unclear speech)

0:03:33.251,0:03:37.491
Try to memorize what she has just said;[br]we'll get to that later.

0:03:38.664,0:03:41.883
With my siblings, and Urit,[br]and people that I get to know,

0:03:41.884,0:03:46.443
I had the chance to see[br]a world full of challenges,

0:03:46.444,0:03:49.454
- a world of special people with needs.

0:03:50.353,0:03:53.772
And this allowed me[br]to examine the existent technology

0:03:53.773,0:03:57.635
in search of an answer[br]for what my siblings were seeking.

0:03:58.542,0:04:02.334
Unfortunately, the current[br]state of the art assistive technology,

0:04:02.335,0:04:07.258
including speech recognition applications,[br]could not provide an answer.

0:04:08.485,0:04:13.534
Since then, all the assistive technology[br]has completely bypassed the voice,

0:04:13.535,0:04:17.411
opting to use other modes of communication

0:04:18.362,0:04:22.361
[by] replacing the voice[br]with symbols and images,

0:04:22.362,0:04:26.222
or movements of the body[br]in the head or in the eyes.

0:04:27.356,0:04:31.806
This brings me to the other lightweight[br]alternative that does use the voice

0:04:32.695,0:04:35.844
which is speech recognition applications.

0:04:35.845,0:04:39.395
This technology works in two approaches.

0:04:40.281,0:04:44.401
The first approach attempts[br]to discover which word has been said.

0:04:46.013,0:04:49.302
The second approach relies on phonemes.

0:04:49.303,0:04:54.443
Phonemes are all the sounds[br]we produce using our mouth and nose.

0:04:55.618,0:04:59.806
Both approaches rely on statistical models

0:04:59.807,0:05:03.136
from a large database of standard speech.

0:05:03.137,0:05:05.959
But once the speech is not standard,

0:05:05.960,0:05:09.659
- when I say not standard,[br]I mean it's enough to have an accent,

0:05:09.660,0:05:11.739
like most of us here -

0:05:11.740,0:05:13.590
this will not work.

0:05:14.444,0:05:19.593
My colleagues and I developed[br]a new approach of assistive technology

0:05:19.594,0:05:22.355
that does use the person's own voice

0:05:22.356,0:05:26.175
and can understand[br]non-standard speech patterns,

0:05:26.176,0:05:31.506
with the mission to give people[br]with a speech disability their voice back.

0:05:32.858,0:05:36.407
So, whose life is this going to change?

0:05:36.408,0:05:39.166
People with cerebral palsy,

0:05:39.167,0:05:41.959
people with Parkinson's,[br]and Myasthenia Gravis,

0:05:41.972,0:05:44.347
so many [other] neurological disorders,

0:05:44.348,0:05:46.637
people who are born[br]with hearing disabilities,

0:05:46.638,0:05:51.717
or people who suddenly have a stroke[br]and their whole life is changed,

0:05:51.718,0:05:54.569
but not only theirs.

0:05:54.570,0:05:58.803
Not only the people who have[br]difficulty expressing themselves,

0:05:58.804,0:06:03.473
but everyone who interacts[br]with them on a daily basis.

0:06:03.474,0:06:08.547
This will make it easier[br]for them to be socially included

0:06:08.548,0:06:13.195
- because every one of us[br]wants to be socially included.

0:06:13.196,0:06:17.508
And now, you may be asking yourself,[br]"How does it work?"

0:06:17.509,0:06:22.078
"How come the current speech recognition[br]technology couldn't do the same?"

0:06:24.978,0:06:27.598
Because our technology works[br]in a different way.

0:06:28.808,0:06:32.217
So, each person has to go[br]through two phases.

0:06:32.218,0:06:35.357
The first phase is called[br]the calibration phase,

0:06:35.358,0:06:41.047
where the person has to teach the device[br]and the application his own patterns

0:06:41.048,0:06:44.227
by entering the patterns[br]and building his own dictionary.

0:06:44.228,0:06:45.920
This phase usually happens

0:06:45.921,0:06:48.920
with the person[br]who understands him the most.

0:06:48.921,0:06:51.090
Together they will build the dictionary.

0:06:51.091,0:06:55.340
This generally takes[br]only one to three hours,

0:06:55.341,0:06:58.280
and it depends on the speaking[br]capability of the speaker.

0:06:58.281,0:07:00.022
After building the dictionary,

0:07:00.023,0:07:03.642
we move to the second phase[br]which is the recognition phase.

0:07:03.643,0:07:07.628
The application will be able to recognize[br]unintelligible speech patterns

0:07:07.629,0:07:10.828
from the dictionary that is already built

0:07:10.829,0:07:14.369
and translate them[br]into a clear voice in real time.

0:07:15.660,0:07:19.819
Our approach is user-dependent[br]and language-independent

0:07:19.820,0:07:23.470
which means it can work[br]in any language in the world,

0:07:24.347,0:07:26.476
even the invented ones.

0:07:26.477,0:07:29.726
And the key word here[br]is 'pattern-matching'.

0:07:29.727,0:07:35.016
Once the person builds his own dictionary,[br]and says a word that already exists there,

0:07:35.017,0:07:36.652
there will be a pattern-matching

0:07:36.653,0:07:39.832
between what he says[br]and what it already exists.

0:07:39.833,0:07:41.852
But here we found a problem.

0:07:41.853,0:07:44.921
We found that people[br]with a speech disability

0:07:44.922,0:07:48.012
pronounce different words in similar ways.

0:07:49.652,0:07:53.601
And the challenge was[br]to differentiate between them.

0:07:53.602,0:07:57.314
So we created a technology[br]called Adaptive Framing.

0:07:58.255,0:08:03.825
Adaptive Framing technology can be adapted[br]to the width of the event in the pattern.

0:08:03.834,0:08:09.543
In the existing technology, you can see[br]the L and the A in the same frame.

0:08:10.402,0:08:15.011
But in our new technology, you can see[br]that the L and A are in different frames

0:08:15.012,0:08:18.042
which increases the accuracy[br]of the pattern-matching.

0:08:18.844,0:08:22.454
And this makes[br]our pattern-matching so much better.

0:08:23.463,0:08:26.352
I suppose you still remember Urit, right?

0:08:26.353,0:08:30.523
Let's listen to her again now,[br]but this time using Talkitt:

0:08:33.520,0:08:34.568
(unclear speech)

0:08:34.570,0:08:36.042
Now I can ...

0:08:36.043,0:08:37.373
(unclear speech)

0:08:37.374,0:08:38.374
... start

0:08:38.375,0:08:39.881
(unclear speech)

0:08:39.881,0:08:41.342
... speaking freely.

0:08:42.982,0:08:44.542
(Applause)

0:08:55.552,0:08:57.906
Talkitt is only one step

0:08:57.907,0:09:02.026
towards bridging the gap[br]between disability and ability

0:09:02.027,0:09:04.946
by letting people express their potential.

0:09:04.947,0:09:07.085
The more we challenge our minds,

0:09:07.086,0:09:11.512
the more gaps will collapse[br]to let us all have a normal life.

0:09:11.513,0:09:12.622
Thank you.

0:09:12.623,0:09:13.653
(Applause)