7pSrWOxsJUc

Alexa ask answer bot question who played
against IBM Watson in jeopardy let me
think about that
Alexa ask answer bought I think pensions
is correct for 35.7% confidence Alexa
ask answer bot question which company
manufactures Gorilla Glass let me think
about that
Alexa ask answer bot I think Courtney is
correct with 100 point 0 percent
confidence now
Alexa ask answer bot question which
worldwide retail chain makes the most
profit let me think about that
Alexa ask answer but I think Walmart is
correct with forty eight point seven
eight seven percent confidence let's get
into the video now
so hello there and welcome to another
tutorial my name is Tammy Bakshi and
this time we're gonna be going over how
you can use the Alexa skills kit with
the Lausanne - IOT platform and IBM
Watson in this case my Aston application
which we'll be getting into in just a
little bit in order to create an
intelligent Alexa application or vast
skill that can answer your natural
language questions getting a little bit
deeper into it let's start off with the
part I'm sure you're most interested in
and that is the Alexa yes that's right
in fact just last week I was in Las
Vegas and I was there for you know for
miles so systems for MIPS conference
while I was there and now we're actually
getting participants as a welcome gift I
was on echo dot and so I got a few as
well now I'm ready and I thought you
know why not create a YouTube video and
how the Alexa skills kit actually works
and how you can use it to develop skills
for your own Alexa device and then of
course publish it to their sort of app
stores skills storing as you could call
it for other people to use as well and
so I thought why not why create a
regular skill when I can combine a lot
of the other technologies that I use in
order to create an even better and even
more useful Alexa skill in fact today
I'm actually taking one of my very first
IBM Watson applications one that's not
only maintained but developed actively
and is in fact open source online github
page called
asked ma asked Anna is an LQA or a
natural language question answering
system that can answer your natural
language questions that have a person
organization or a location answer type
it uses the power of Ivan Watson natural
language classifier and natural language
understanding in its back-end as well as
bi-directional attention to flow by
Allen AI Institute
alright now let's get a little bit
deeper into how the actual Alexa skill
works and how this new topic of the
Lausanne to IOT platform actually ties
in with it let's take a look all right
now getting a little bit deeper into it
let's start off with taking a look at
the flow of the application now of
course we have the Alexa itself in this
case I've got a doc device but any Alexa
will work you could even use a simulator
online you could use Amazon's official
Alexa simulator all these devices will
work perfectly for this now once you've
gotten your Alexa device up and running
and you're ready to start off what you
need to do is sign up for a platform
called a losan to IOT platform now when
you're designing an Alexa app you have
to but you have to wait two methods to
go down two rows to go down either you
can use AWS lambda which is what's
recommended by Amazon and if you use
lambda which is Amazon server less
compute platform then what happens is
you're going to be using this and of
course I was on Web Services I'm just in
case you are wondering Amazon all-access
skills kit is not part of the Amazon Web
Services they are entirely separate
they're part of the Amazon developer
console suite and so in this case though
I didn't want how to take you into the
world mammals on Web Services just yet
let's kiss today let's actually try and
develop something with a REST API to
keep it local on our own server and so
what I've done is actually I guess you
could say a different way of creating an
Alexa skill and the lausanne to iot
platform is a very easy way of doing
this because what happens is through the
lausanne at the Lausanne DMT platform I
don't need to do any manual
authorization or the SSL Certificates or
anything else that Amazon requires you
to do when you're working with this
technology and so using Lausanne
I actually opened up an HTTP endpoint
and this HTTP endpoint can be contacted
via Alexa we'll talk about that in just
a moment now once we've gotten that
Lausanne to IOT platform up and right
to take a look at what exactly those
avila Laur alexa to communicate with in
this case we're actually communicating
with of course ask Tanmay which is my
cognitive application built using IBM
Watson we'll get to that in just a
little bit actually
there's an entire suite of applications
that we're going to be using and they're
built in Python
so let's observe this section for Python
now we've also got a flask so this is
actually building these flask API this
is a flask HTTP server again apart from
that we have a few more modules in
Python including this answer cache and
I'll talk about what that is in just a
second all right now these are the three
main components inside of our Python
tools suite here and let's take a look
at how all of these actually incorporate
together however the thing is this is
running locally on my system and why
mess around with port forwarding but we
could just use an croc or ngr okay and
so of course in the cloud here we've got
n G R and the Jeremy's gonna allow
Lausanne to communicate with my flask
HTTP server now let's take a look into
this now one more thing I would like to
say here because I gonna say a huge
thank you to market suitor macht who is
actually a developer advocate at IBM for
actually Helen you build and fix a few
issues with the Alexa skills kit so the
thing is when you're working with a
stamen you want to ask or find the
question of the user is asking the
system in order to do that you not just
want an intent you want the fact you
want to raw utterance what the user set
unfortunately Amazon does
support this however using a workaround
that marks help me actually figure out
using slots or as we'll be talking about
in the actual programming part we've
been able to extract that raw utterance
and of course using that I can actually
send that to over to a standby and then
ask ten may can give the answer back a
good feeling all right let's take a
little bit of a deeper look now now of
course when you ask the Alexa a question
what's happening is let's just say you
say I like so ask and in this case the
invocation name for the application is
answer box okay so the this is the
invocation name and then you say
question and let's just say your
question is ABC doesn't matter what it
is okay now this is what you tell your
Alexa now once you give your electorate
questions instead of actually having
this quote here let's just have
questions of variable sleeping zoos a
little bit better and as a sort of
diagram here so let's just say you've
got your question so you can ask the
Alexa the question and you've told it
that you want this question to go over
to the answer bot invocation name
now once you've told that the Alexa is
going to try and figure out what to do
in this case it's going to go ahead and
contact the Lausanne HTTP server
now what Lausanne is gonna figure out is
actually find out what you're asking are
you asking a new question or are you
trying to figure out the answer to a
question you've asked previously I'll
talk about why this is important in just
a moment now I'll just pretend of course
we take this question through Lausanne
and on Luzon also sends over to the
flask HTTP because remember in this
flask HTTP HTTP we've got an endpoint
called Alexis
all right this is our end point over
here this is the HTTP endpoint that
little Santa's gonna call through ng on
okay and some goes through in gr okay
and of course comes over to the Alexa
skill endpoint now once it's at this
endpoint the interesting stuff starts to
happen see this endpoint is actually
going to contact the a stem based system
which itself is an isolated system now
as Tanmay is going to store the answer
in an answer - now why is this you may
ask you see asked an mate is efficient
in fact I've actually rewritten as Tammy
from scratch and Python to make it as
efficient as possible
it's even more a cure than the four as
well however it's not efficient enough
for Alexa to answer instantly because of
this Alexa doesn't really work out with
asked an me immediately and this is what
you need to do is tell Alexa the
question Alexa will tell you that it's
thinking of the answer and two or three
seconds later you can ask Alexa for that
answer because Alexa expects an
immediate reply from the as ten main
HTTP endpoint in fact let's see a demo
of this in action now of course this is
on mute so it's actually unmute our
Alexa and take a look at how the
application works Alexa ask answer bot
question which company is a long Musk
the CEO of let me think about that
alright so as you can see Alexa is now
thinking about the answer to this
question and now in theory if the speech
recognition actually worked out nicely
what I should be able to do is this
Alexa asked dance Alexa asked answer but
I think Tesla Motors is correct with 100
point zero percent confidence there we
go that's a lot of confidence and so as
you can see I asked this Alexa which
company a lot of musk was the CEO of and
at guest Tesla Motors with a hundred
percent confidence I look at something
Deaver shall we Alexa ask answer bot
question who is the CTO of IBM Watson
let me think about that
all right it's not the Alexa is actually
sent the request or an ass hand meant
through the Lausanne Taiji platform and
it is now processing now for around two
to three seconds processing we can do
this
Alexa asked answer but I think Rob PI is
correct with fifty eight point two one
two percent confidence there we go as
you can see it says raw high as the
correct answer and Rob high of course my
mentor is indeed the correct answer
here's a CTO now though let's take a
look at one last quick demo on the mill
get back over to the actual architecture
of the system itself let's take a look
Alexa ask answer bot question which
company is the largest fast-food chain
in the world let me think about that all
right Alexis thinking again it's sent
back request over to ask tan may ask him
is doing the old processing using IBM
Watson and now Alexa asked answer but I
think McDonald is correct with ninety
six point six to one percent confidence
that went very smoothly and so as you
can see as Tammy was able to give us the
correct answer and it's to this super
portable super convenient Alexa
interface anyway back to the
architecture hope you enjoyed that demo
you'll be seeing a lot more in the
programming part all right so now of
course we've got this entire system and
you now understand why the answer cache
is important and this is also why we
can't just have one individual endpoint
we can't just have the Alexa skill
endpoint we need to have one extra now
this extra endpoint is going to allow
this lausanne HCP to actually grab but
the answer once asked Tammy is done
processing now this endpoint is called
slash yet
answer alright
in fact gel so you have to see a little
bit of a clearer picture of what these
endpoints actually do let's list down
the endpoints over here now of course at
first they are HTTP endpoints we've got
the Alexa skill endpoint all right now /
Alexa skill now the parameter so this is
going to take in first of all it's going
to take the question and this is the
question that the user asks number two
it's going to take the user ID now this
is actually just user in the rest call
and users a unique user identification
number that Amazon gives us it's a very
long string that actually gives us the
ID of this user it's entirely unique to
that user and we use this to store their
answer in the answer cache without you
can restore the question which other
people may have asked as well alright
now after that though we of course do
have no more end point and an end point
is the slash get answer HTTP endpoint in
terms of parameters we take only one
parameter and this is the user ID
parameter and the reason we take this
parameter of course is to find out what
the users answer is and it returns a
string which is exactly what the Alexa
needs to say out you know what it said I
think MacDonald is the correct answer
with such percent confidence that is
exactly the string that get answer
actually outputs and that's what a
steaming computes and so this string
will go back over Lausanne and into
Alexa Lake south
so now imagine this you've asked your
question to Alex already but now when
you want to retrieve the answer to your
question so you send the retrieve
command all right so you sent the
retrieve command and this is of course
with our red to get answer end point and
so once you send this over to your
Lausanne to HTTP what's going to happen
is again through ngr okay lo Santa's
going to go over to the get answer
once it's applicant answer endpoint is
again world of fun starts to come in
because what's going to happen now is
it's going to go over into the answer
cache and it's going to ask for that and
for that user IDs answer the answer
cache is going to send a response and
this time what's going to happen is
through ng R okay it's going to respond
over to lo sont and the Mossad is going
to send this final response and let's
just say this goes over through Lausanne
it's going to send it over to oral exam
and it's going to tell their likes of
exactly what it needs to say and this is
one of the architecture diagram of what
all truly happens see the best part of
using the Lassonde as you platform once
more is that we need to do almost no
code at all for the alexis kill the only
real code here is this part this part in
this part that's it everything that's
not Python requires no code and it's not
entirely and a graphical user interface
whether it be with the Alexa skills kit
console or development environment or
whether it be the Lausanne - IOT
platform their graphical user interface
and gr ok isn't really anything you need
to program with anyway and then of
course under the Python flask issues if
you relatively simple flask server I'll
be showing you the code behind that
just a minute the ants will catch again
very very simple code and asked a mate
that's where the real meat comes in
itself does have quite a few different
steps what it's able to do is actually
use a search engine in this case it uses
the Google Search API and what it does
is it doesn't actually extract pages it
doesn't extract four pages from which to
find answers
it only extracts summaries what do I
mean by summaries well it actually is a
step called search and results summary
extraction and you know when you google
something and you see those little
descriptions under the little links that
show the parts of the webpage that are
most important to your query that is
what asked and may actually parses it's
going to grab all those from the API and
it's going to use a bi-directional
attention flow from a Linnea I would buy
own sort of modifications to that and
what it's going to do is try and find
some kind of answers to it it's also
going to use IBM Watson natural language
classifier and to find out what kind of
response are looking for whether it be a
person organization or location once it
knows what they're looking for it'll
filter out according to that using the
natural language understanding API to
find out what all the different answers
are and then once it's done that's what
you do a little bit of naive answer
scoring try and score based off a
repetition and if it's also going to
score based off a PageRank score so you
know that Google uses the PageRank
algorithm along with now a different
deep learning algorithms but it's going
to use that score in order to rank
better answers that are towards the top
of the list with more confidence and
that is how the entire ask 10ms system
works of course as Tanmay and how it
works is an entire video in itself but
in essence it's able to use a
combination of tensorflow
Watson natural language understanding
and what's a natural language classifier
in order to get answers from the web
from your natural language questions so
that's a quick brief of a Stan there
what that does and everything else
including the Alexa skill the Lausanne
platform and John okay the flask HTTP
server
and the answer - Tammi work together and
how they actually enable me to ask this
Alexa and natural language question and
get an answer around 95% of the time
alright so I hope you enjoyed that quick
demo and of course the architecture
explanation of how exactly this all
works now though what I'd like to do is
I'd like to take you over to the
programming part where I'm going to show
you how you can actually build a skill
for this Alexa and I'm gonna take you
through how exactly I built this entire
system how you take the wrong uh turns
from the Alexa skills over alright let's
get into the programming part now
alright so welcome back and now let's
take a look at how you can actually
implement this Alexa system alright now
what I'd like to do first of all is show
you what's actually going on in the back
end now I've actually got my Alexa here
with me
and what I'm gonna do is I'm gonna turn
this on unmute it and I'm gonna go over
to a four so you can see the Amazon
Alexa skills get here but I'm also going
to go over to these two terminal windows
these eternal windows contain of course
the ng ROK server is actually running on
our custom sub domain actually and of
course on this side over here I've got
my Python server which is hosting the
flask HTTP endpoints of get answer and
Alexa skill and of course it's also
hosting the answer cache and contacts as
Tanmay in fact an entirely new
redesigned version of a statement as
well all right let's take a look now
let's actually I want to run a quick
demo on the actual device and let's take
a look at how these two terminal widows
actually react let's ask it a simple
question like just for example Alexa ask
answer boffed question who is the CEO of
Amazon let me think about that alright
as you can see the Alexis now thinking
about it and Angra has received the HTTP
request it returned 200 ok response and
you can actually see that asked Anna
over here has actually received a
question who is the CEO of Amazon
person who actually sucked this is this
user ID and as you can see asked him
extracted the answer i think jeff bezos
etcetera and it said that this answer is
for user ID etc etc in fact we can
actually now go back over to our Alexa
and say Alexa ask answer bot
I think Jeff Bezos is correct with
seventy two point six to two percent
confidence alright so as you can see the
Alexa was able to detect that or not the
Alexa the election was able to send our
user ID over to asked and may ask Tammy
is get answer was actually able to take
that user ID and I was able to extract
the correct answer associated to that
user IDs question and that is how the
system works in fact you can even see
that certified right and you're okay
over here who is actually getting those
requests but wait even ngr okay has to
receive its requests from loccent
so let's take a look a little bit deeper
into how the system is built in fact I'm
actually going to link to a guide on the
low sound blog that's how you can build
an interactive Alexa skill with no code
using this lausanne alright so this is
going to be in the description you can
actually take a look at this tutorial is
going to be very helpful even actually
where you're breathing when you're
building this application in this
tutorial I'd recommend you take a look
at this blog as well just to find out a
bit more about lausanne - how it works
and what it's meant to do in fact to say
you know this is actually the arrest API
that Alexis till end point is actually
calling just for example if I were to go
localhost 5000 which is the port it's
running on go to the Alexis skill
endpoint pass the question in this case
who is the CEO of Apple and give it any
random user name like ABC I run that go
back over here as you can see a Stanley
has received the question and the
username ask Tammy is now going to reply
and as you can see Google Chrome gets
the response back and now this is what
actually goes back to Lausanne well
we're back into the answer cache and
then when you ask for the answer
Lausanne is going to receive from the
answer - and give it back to you like so
all right but let's start from the
basics here let's start from the Amazon
Alexis
Gill's k now the Amazon Alexa skills kit
we can build chat BOTS and other kinds
of skills for the Alexa now I like the
new console that they built but you can
use the old one as well if you'd like to
at least for the time being there we go
alright so as you can see these are all
my Amazon Alexa skills you know Tenma QA
and asked handmade v2 that's Tanmay so
it's actually go ahead and click on edit
now when I click on edit you can see all
the stuff that goes into this
application let's start with the
interaction model now the interaction
model has something called an invocation
an invocation is how the user actually
begins and talks to Alexa like just for
example let's just say your invocation
name was daily horoscopes you could say
something like Alexa asked daily
horoscopes with a horoscope for Jen and
I which is the example that Amazon gives
us here in this case my invocation name
is answer bot so people can say Alexa
ask answer bot question who is the CEO
of IBM where Amazon or any other kind of
question that you may have there are a
few requirements to your invocation name
like you cannot have phrases like launch
ask tell load begin our naval or even
wake words like Alexa Amazon echo
computer or the word skilar app so these
are the different sort of requirements
for your a vocation name but that's all
once you've got a good invocation name
and you're ready to continue we can
start off by creating a slot type now
before we actually create a slot type
though we have to understand intense if
you've ever used the IBM Watson
conversations in this before or for that
matter API today yeah I or any other
kind of chat platform you're familiar
with these intents but you may not be
familiar with slots so let's talk about
that now inside of the intents we only
have one intent and it's called the
everything intent remember this chat bot
does nothing the only thing that really
does something is asked and a so you
want to take what ever the user says no
matter what it may be and send it over
to lausanne and lausanne can do the
actual post processing and understand
what the user actually bent so with
everything intent we basically try and
capture everything that the user says
now this includes no matter what they
say they can say question they can say
everything intent but then after that
whatever they say after the first word
will be considered part of the
everything slot and the everything slot
is meant to pick up whatever the user
sets doesn't matter what it is and
everything slot as you can see over here
has under the intent slot is actually a
specific slot type the slot type is bag
of words now this is a custom slot type
Animas on doesn't provide this by
default and so what you need to do is
create a custom slot type this slot type
you can give it any value in this case
we've given at hello world unless you as
a user specify something else in your
intent which is what you should do now
of course you don't need to give us any
meaningful intent or meaningful value
you just need something in order to be
able to capture everything and send it
over to IBM Watson in fact if you take a
look at the IBM code pattern for the
Alexa you can take a look over here that
as I was mentioning just wait for this
to load for a second Marc Stuart Avant
and Nicholas haidle off worked on this
code pattern together and they use I of
the Amazon Alexa and IBM Watson in order
to discuss the weather and milla
conversations or really just choose one
from a library and so this is actually
based off of IBM's first of open wisk so
this is of course IBM's sort of us
compute platform similar to AWS lambda
and so of course this code pattern is
basically it's quite similar to my code
to my code in the way how it actually
allows Alexa to communicate with the IBM
cloud functions or in this case about
the idea functions but Lausanne to IOT
platform alright I'll tell a little bit
deeper into this now you know how the
intents and the slots work once you
built your intention your slots all you
need to do is save and build your model
once you've saved and built your model
though you are ready to actually test
your application now this is where you
can actually test out
your Amazon Alexa app before that though
let's head over to Lausanne now as I
mentioned no code involved however if
you want put some code in and get a
little bit more flexible there you can
and that's what I've gone ahead and done
but this is what the Lausanne flow looks
like and so as you can see I can just
format a little bit here as you can see
it's a very sort of graphical user
interface similar to IBM IOT ZnO to read
which is open source I wanna close on
however basically what I'm doing with
those on here
and I'm creating a web hook now this wet
book is basically the HTTP endpoint that
the Alexa skills kid calls how does it
call it well if you go over to build
over here if you just wait for one
second as you can see there is a button
for the endpoint and of course you can
actually get that over here as well if
you click on endpoint you can either
choose AWS laptop or an HTTP server in
this case I've given it to my Lausanne
trigger and of course you to choose if
this my development endpoint is so
domain of the domain that has a wildcard
certificate term a certificate authority
doesn't matter
that's just what Lausanne tells us does
choose because that's how Lausanne
destructure
anyway once the web hook receives
information from the alexis skills kit
then though we actually call a custom
function now this is a function coded in
JavaScript that I've created it's a very
simple function all this function does
is it actually checks whether the intent
is to actually ask a question or if the
intent is to find the answer to a
question that they had asked before if
they want to find the answer to a
question they've already asked then Dana
dot works is at the zero or else it's at
to one if they actually want to ask a
question once that function actually
goes through there's a conditional this
conditional is whether or not they
actually want to ask a question if they
do want to ask question that it goes
down through this flow in this flow we
actually respond to the alexa and we say
let me think about that which is why
alexa says let me think about that when
you ask
question and then over here with the
HTTP what happens is your question is
sent over to that mg ROK server and of
course the question is passed and the
question in this case is if I make this
a little bit bigger data body dot
request on intents lotsa everything slot
that value if you go huland JSON that
the Alexa skills kit provides in the web
hook this is exactly what it looks like
and then the user ID is given over here
which allows of course asked me to keep
its answer cache all right so once asked
M has gotten the request and it starts
doing its own processing and of course
Alexa's already told you that it's it's
it's thinking about what the actual
response should be and so now two or
three seconds later when you ask Alexa
to retrieve and when you ask Alexa to
actually get the answer to your question
then a the conditional goes over through
this branch now in the conditional ghost
of this branch it still does make an
HTTP request in this case the request
goes to a slightly different end point
they get answer end point and this time
not the question but only the user ID is
sent now once you actually receive a
response from the HTTP endpoint in fact
if you were to take a look here it
stores that result in data to ask Tammy
and then I've got a a custom JavaScript
function and what this does is it
actually creates an Alexa response now
if you remember over here we sent a
reply to Alexa and we sent some JSON and
the text was let me think about it in
this function what I do is I create a
similar JSON variable but in this case
the text is payload data asked and made
up body
now remember asked animate is the result
from the asked anime endpoint and body
is what asked him a returned and so
we're basically taking whatever asked an
air turned and we are basically we are
essentially giving it over to Alexa and
then of course we reply with this Alexa
response over to the Alexa and again
since this happens practically instantly
just a little bit of latency it's it's
good enough for Alexa to to not say
the request timed out or anything that
anything of that sort but now though
let's see a little bit of a deeper dive
into what Lausanne to seize from its end
so now what we can actually do is find a
few debug notes all right now these
debug notes allow us to actually print
out what the system is seeing or
basically doing at each step so we can
actually print out the result from the
actual web hook and this the web hook
debug and let's just message this as log
for web hook all right now when this
when when the web hook when the web
hooks results printed it's going to tell
us that this is the log for the web walk
web look we can do a similar debug for
for example let's just say over here we
can take a similar debug after this
function all right and then after this
function we can take a look at what the
actual result was final function debug
or lock and then what's to put one more
debug here why not and let's connect
this to HTTP there we go all right so
now we've got our D bugs in the right
places let's just say asked and I'm
going to deploy this workflow to figure
out there all right so just deploy this
workflow and as you can see the workflow
has been saved and deployed now if we
were to go ahead and ask the Alexa
another question it should actually
print out everything in lausanne if we
go over to our debug I'll just replace
remove all the previous locks and let's
go over into our alexis field yes go
into the testing and now what I'm gonna
do is test out the Alexa now either I
can talk to my computer and it'll
simulate the Alexa or I can just type it
in let's try typing it in and then next
we'll try actually um saying it so it
will say ask answer bot question
found it let's just say Amazon why not
I'm gonna send this over as you can see
Alexa replies let me think about that of
course that's exactly what we expect and
that's what lausanne replied but let's
take a look at what all lausanne to
actually saw as you can see we have two
different logs here why not three
because this was never called now the
first one is the log for webhook
as you can see Orion as the web hook was
actually triggered this is all the data
that we had in our payload which is the
sort of group of information that we've
got in this entire workflow now data is
where we actually try or is where we
actually start to get all the data that
came from the web hook now inside data
we've got the body and this is the
actual information that the Alexa sent
over now inside of request you can see
the intent the intent of the attain
group over here and the name of this
intent is everything intent and there's
no confirmation now under slots however
and under everything slot you can see
that the values who founded Amazon
sounds like a little bit of a workaround
because we can't get the raw utterance
so it is but until Amazon actually
provides us the functionality you can
get the raw utterance we won't be able
to use anything but this kind of
workaround but it's a very good
workaround nonetheless
and of course again huge thank you to
mark for actually helping me implement
this as you can see of course they're
resolutions that doesn't matter all that
really matters is the value and since
the application realized that the value
is indeed a valid question it sent this
over through this line instead of this
line which meant that as ten may
actually received that question and it's
now processing in fact if we go over
here you can see that ng ROK received it
and as you can see asked how they
received it here as well and it has
created the earth has thought of the
answer that has replied and sorted in
the answer cache of course all right so
after the log for web hook there's also
the asked and a result log because
remember when you actually call slash
Alexis kill endpoint we actually don't
return anything
the Alexa but the end point itself does
return some JSON and so that JSON is
never actually saved in the I guess you
could say in the payload however it D is
actually replied nonetheless alright
next though let's go back over to our
alexa simulator and let's try actually
using the the speaking feature ask
answer bot question retrieve okay so as
you can see Alexa's voice voice
recognition hasn't really worked out
that well due to the fact that my
microphone setup isn't optimal right now
it's a little a little funky but if I
were to go ahead and type in our
question once more as you can see it's
thinking about that once more oh we can
go back here in fact let's go over to
Lausanne and you delete all of our logs
because we don't want all that noise in
there let's go back over here and say
ask answer bot question retrieve don't
worry let's try that once more all right
as you can see Alexa does in fact
respond with I think Jeff Bezos is
correct in forty six point six three
eight percent confidence and that is
indeed correct Jeff Bezos founded Amazon
now of course if we were to go back to
Lausanne you know that this time of
course a log for web hooks he'll Silla
actually went through which is why this
is printed out but after that this
conditional went through this branch
instead of this one and so through this
branch the function goes down and we
actually print out the result as well
now if you take a look in the final
function log we have of course our
regular body and everything but we also
have to ask Tanmay response and this is
what asked and they responded from the
get answer endpoint and then we've got
the Alexa response and this is exactly
the JSON that is sent over to the Alexa
and then the Alexa is able to make sense
of it and actually respond with this
text and that is how the lausanne to IOT
platform ties in with this entire system
in general this is all able to combine
in order to
to create an Alexis skill that uses the
power of ask Tanmay in order to answer
your natural language questions and so
that sums up how you can build an Amazon
Alexis skill and not just any skill in
fact quite a complex skill but with
hardly any coding involved only answer
cash the asset and a system flattens all
you need to code everything else on the
Alexis skill side is all done without
code whatsoever I really do hope you
enjoyed this tutorial and that you're
able to create your own I like some
skills as well of course thank you very
much everyone for joining in today if
you have any more questions or
suggestions or even feedback please do
leave it down in the comment section
below email it to me at a g-man you to
mental comb or tweet it to me at Tai Chi
mani or axe had you met me on Twitter
apart from that though if you do like my
content you do want to see more of it
please do consider sharing liking this
video and of course subscribe my youtube
channel is it really does help out a lot
of course turning the notifications if
you liked me notify and whenever I
release a new video
again source code for this application
it'll be available down in the
description below ask 10-man will be
updated on github very soon as well
after some minor code cleanup within the
next few days you'll be seeing an
entirely we were in version of asked him
they on its github page where it belongs
alright again thank you very much
everyone for joining in today that's
gonna be all for this tutorial
hope you enjoyed goodbye