Press "Enter" to skip to content

The future of voice in the workspace. | Srikar Yekollu | TEDxBangaloreSalon


now whose job is to proceedings and
transcribe them and why is this
important it’s because this makes it
very efficient ways to make sense that
technology into this picture and that’s
essentially what happened now one
and dumping these technologies one way
to do that is to work backwards right
what is language composed of nights when
I move this composite of birds and
smallest units of language which make a
semantic sense and the process starting
from all your signals and working
backwards towards these worlds is called
transcription and this is an important
and difficult problem in Excel and this
long time this German is a long arduous
one starting from being able to
recognize Bizet’s action it’s just the
system of water systems now this journey
has been bigger so these X
transformation systems of is such an
extreme bed for certain language accent
pairs we can call this problem salt
for example the key is language exit
rate this is languages like us English
and the rest is on the remaining
accidents remaining set of languages now
salt that there is a measure which can
recognize this a human parity the best
way to human can do
there’s the red machines which can to
exert at the same level on data sets
that’s what salt means here now well
next what makes for these kind of
systems what’s next is that we need to
work in terms of being able to make
sense of what these words mean right
then put together levels of complexity
what do sentences mean what do these
sentences put into paragraphs mean what
do these when these conversations mean
and that’s the next step and that’s
essentially where a lot of work is
engineering efforts and that is what my
and now let’s see you’re resetting
agents and this process to make sure
your agents are delivering the right
quality of service now how this is done
is typically with sample or small set of
pearls and you have people listen to
these calls and make sure is going as
expected and this is a very
time-consuming process between 6 to 10
minutes to edit the skull we believe
that if the time and you don’t have to
deal with sampling of any of this
another example of this in the this to
the scenes domain now if I ask you what
makes a good sales agent one who talks
moved in the customer or one who talks
less than the customer so my intuition
when is this works that as a personal
sales agent will actually be talking
more than the customer trying to make
this point trying to sell everything
that say open that has to say about the
company that doesn’t sadistic don’t seem
to say so a good sales agent is he
now this kind of insight and this is
some this is something which is not only
good for them there is something more
actionable for them and these things
that transform workplaces so RV as
I can tell you and and and things like
that and these first electronic computer
this is a ballistics research human
bring us to do and ultimately since time
we’re actually very big and can you think machines can help
Please follow and like us: