fUL93_G3yuI

so long there and welcome to another
tutorial my name is Tina Bakshi and this
time we're gonna be going over how you
can use the IBM Watson conversation
service in bluemix with a special
surprise actually just like the natural
language classifier service the
retrieving write service and you know
the dialogue service for example they
all have sort of toolkits that you can
use in order to easily train your IBM
Watson service and so now as you can see
the classifier I mean the conversation
service as well has that little tool box
where you can easily trance to do what
you want alright so now let's get right
into it so now first of all what is the
conversation service now first of all
the conversation service is essentially
a mash-up of the IBM Watson dialogue and
the IBM Watson natural language
classifier services now if you've worked
with these two services before you know
that combining them can create really
interesting chat BOTS I guess you could
call them so someone actually even
created a movie assistant using the IBM
Watson dialogue and natural language
classifier services and so basically
that's that's really a neat application
actually it essentially allows you to
you know say I want this this sort of
movie I want it to be rated this I want
sort of comedy and it'll be able to
create this little list of movies that
you'd like to see but today I'm going to
be showing you how you can create an
application with a conversation service
which it was essentially this chat bot
that allows you to order pizza you can
give it your specifications like you
want large or you want no sauce or you
want these toppings and then it will be
able to actually order the pizza for you
I guess all right let's get straight
into it now first of all before I get
into this what is the conversation
service do it allows you to create
applications for users to be able to
converse with IBM Watson and so how does
it do this now let's get into this now
there are three main things that go into
making this work let's begin with number
one then there's number two
of course and oh sorry how about that
and of course number three okay that's
really number one what is number one
it's intense
now what is an intent in this case are
usually by definition it is what a user
will intend to do basically a user's
intent okay so this is basically the
waste whatever message that the user has
given IBM Watson or the conversation
service in this case it's going to say
what was their intent behind that
message it'll basically use machine
learning this is where actually the
natural language classifier comes in and
this is where basically it finds out
that hey the user wants to do this for
example if I say hey what sizes of
pizzas are there okay and it'll respond
to me small medium large and extra-large
however behind there it knew that hey
his input meant that his intent is to
find out what sizes of pizza there are
all right once it understands the intent
it can actually reply to the user
next entities now what is an entity in
this case we're essentially scanning the
users input for uh Nanta t so basically
entities in the users input okay so we
basically take the users input and we
filter down to look for entities inside
of that statement
now let's say for example I were to take
this let's say why I said you know top
topic number one should be cheese all
right and so basically what it's going
to do is it's going to say hey the
intent was to set a topping to cheese
but the entity was
what was the number-one meaning we want
to set a topping the cheese and that
topping is the first topic which is how
entities will tie into intense so that
the conversation service can really
understand what the user wants and you
know using these two combined throughout
it throughout an entire conversation you
can have contextual analysis as well
so if you say hey my first topic should
be cheese you can say hey next I want
chicken for example or pineapple okay
and this will actually come into use in
a real application but not in this video
though I will be showing you a demo and
a quick tutorial on how to create your
own applications using the conversation
service today I'm not going to be going
into how you can actually use the REST
API to create your own applications
that's for another part today I'm going
to be showing you how to use the toolkit
to train that conversation service so
now finally the third and one of the
most important sort of basic aspects to
this so now first this is actually the
dialogue step now the dialogue step is
essentially the conversation there's a
second
it's essentially what I like to call the
conversation flow all right it's now
basically well the eraser doesn't want
to stay but basically what this is is
the dialogue aspect to this is where the
dialogue service comes in and so
essentially as you can see what happens
here
is it essentially creates the
conversation flow it is a set of
conditionals that says you know what if
they specify that hey I want to add a
topping that topping is cheese and they
said it's the second topping that they
want to add you can wash those together
and - you know what they want to add
cheese is their second topping onto the
pizza okay and then you can respond to
them and you can also check hey they
didn't they mentioned that they wanted a
topping his cheese but we've done the
textual analysis and we have also you
know I gone through the entities and we
didn't find which topping they like that
to be and so we can respond to the user
by saying you know what you haven't
really told us which top you'd like to
have to be so we can't add that to your
pizza please tell us whether they would
like that to be topping one two or three
and that's how the dialogue aspect can
tie in with intents and entities in
order to create a applications that you
can converse with just like a normal
human
alright so now that's gonna be it for
this little explanation part so now
again this is going to be an application
a chatbot for you to be able to talk
with IBM Watson just like a normal human
being so that you can actually do
something like order a pizza then again
this can be applied to many different
things TVs cars phones anything all
right let's get to the Mac part where I
am going to be explaining the actual
code and a little demonstration to you
so welcome back to the Mac part and now
I'm going to be showing you how exactly
you can use the conversation service to
make a pizza man that can take your
pizza order and yeah send it to the
person who makes it and yeah and then
they can have a pizza so let's begin now
first of all what you want
do is actually create a conversation
service and Linux so to begin what you
need to do is go over to bluemix first
of all and go to your catalog all right
then go down to of course lots and over
here watson then click on conversation
your plan i'm just going to go for its
41 for now i'm going to put it in a cell
in any space really make sure you keep
track of the service name though the
credential name really whatever again
select the plan is free and you should
usually click create but i've already
created this so let's go to my dashboard
and as you can see the service i'm going
to be using today is conversation gk so
alright so now what i'm going to do here
as you can see this is the conversation
sort of dashboard i guess this is where
you can go to the developer resources
this is the introduction page so now
what i'm going to do here is i'm going
to click on the launch tool button ok
this is going to launch a tool that will
allow you to sort of train a watson
conversation service
now again you want to make sure that you
will recent the correct instance of
conversation since you may have many
instances as you can see in fear you
create something called a workspace
which is where you're actually going to
be creating let's say your pizza mat so
let's just say create a workspace and
i'm going to call this it's all mine for
you too all right after so watson pizza
all rights now english glass and let's
just create it as you can see it i just
created this little workspace here and
if i go to get started
it allows to put things like intense
entities and dialogue now we do actually
to go back to this other workspace which
I have actually filled out so I don't do
you know go through filling it out which
would be extremely long basically as you
can see here in the intense page I
filled out all these intents such as
what capable for example are which is
basically how can I customize my feeds
are what are the options what can you
change or you do what options do I have
what you have to order it's not socially
what this is doing is all of these
questions that I've given it leaves out
this list are these cells these six
questions what happens is these actually
go to the natural line it's classifier
service and say you know what these
questions actually if you trained to it
I mean if something similar to these
questions arise then what happens is we
know that the what capable intent is
what the user is intending to do or
intending to look for it's that sort of
what's hot what's happening to the
natural language classifier how that
ties it and again this is very useful
here because the thing is before you'd
have to actually manually program in the
combination of the natural language
classifier and dialogue now it doesn't
for you so as you can see in these what
capable what's happening is as I said I
mean then dialogue over to this dialogue
here why does the user asks something
that is about what cable it says you
know I can help you customize your pizza
and following Roy's size toppings
crossed thickness and sauce all right
slava go back there are they're intense
as well like the topping
cheese okay a cross thick sauce extra
sauce normal sauce no and again all
these intents are essentially what the
user could want to do here so for
example let's say we wanted to add an
intent that allowed you to ask a
question like when where did you two get
here so let's say we said one um pizza
bit of a time okay well let's just say
delivered all right asana well let's say
these are some sort of sample questions
of these are class how long will it take
to arrive when will the pizza very good
ten minutes is it free after a certain
amount of time also for example um I'm
patient for pizza all right so now these
are sort of examples of what the user
could ask in order to have an intent of
delivery time so now if I click on done
as you can see this intent has been
added and so now I can go to my entities
again I'll get into actual programming
the intensity dialog in just a minute
but in the entities we only have the
topic number which is the first second
or third
as you can see sort of which topic
number were looking for let's every want
to add a fourth one actually all I need
to do is go to topic number add a new
value okay and say you know what let's
say you want to add four okay now Riley
report so then I can add some synonyms
or other ways that user could mention
that entity so I could say four number
four or or or enter or number four or
fourth or four or topping for right
those are some ways or you know just
yeah
and so let's distribute some ways that
the user could mention topping number uh
and so it's like you know just click on
this little plus button over here it
adds the fourth topping capability here
but then again I only want stick with
three so if I just put this little
checkbox here and if I click on delete
they will delete the fort topping them
if I want you to create another entity
in general just click this little crate
button on top here now let's say we go
for a dialog all right now I mean we
have incorporated this new air intent
call delivery time and then
is when the user will be able to get
their delivery now if we go to the last
dialogue over here we click on this
little plus button down there as you can
see I can just minimize that now now as
you can see here I can enter a condition
when this dialogue is triggered by
Watson so they these are asks you know
delivery time right I just click on off
then what we can say the pizza should
arrive within 30 minutes or it's free
all right then just minimize that
that'll also save it and now we are
ready to test out the conversation
service now as you can see if we go to
this top right over here you click on
this little conversation icon
as you can see Watson so training on our
recent changes that actually gives me
some time to go over the previous
dialogue that I've already created
let's begin first of all if the user
asks for small pizza all right a small
lar medium then already media same thing
to the other sizes and show you gets you
catch the drift and if let's say they
ask for no sauce then no songs its ask
for normal sauce general the default
sauce is ask for extra sauce extra sauce
it is if they want thin crust
sounds good thin crust if they want
thick crust the question will be however
if they want cheese on a pizza
and they've specified which topping we'd
like that to be then there whatever
topping later that to be is cheese if
they same thing for chicken all lives
and pineapple but if none of these will
trigger and they've fallen through just
like that else if statement if they've
fallen through each one then as you can
see it falls over to this one where no
matter which topping is selected you
know that they haven't specified a
specific topic number and so we can say
you know what specify which topping
number you'd like this to be next we can
say what I mean we if the
what if the user wants to know what
Watson is capable of then this is the
answer to that I can help you with the
size toppings cross thickness and sauce
you can of course answer what sizes
there are of pizzas it can answer what
toppings you can put on your pizza we
can answer what thicknesses there are
I'll crust and it can answer what types
of soft starters
then we can say you know what there's
this also intent where a work complete
the conversation and so the user said
you know what sounds good let's go to
the next step I'll tilt wheeling billing
details so we say Watson says sounds
great this receipts the next page where
you can search on a price and size
shipping address and the details it's a
lot some pizza this large up there in 30
minutes it's free just and if there's
anything else then you know that's
usually the first message that we're
sending to the user so just going to say
hi welcome to Watson pizza
I'm Watson may I take your order all
right so as you can see on the side here
Watson is actually done training it
tells us so now we can actually you know
start treatment start you know testing
out what all the conversation service
let's do that shall we
now first of all it says hi welcome to
our sweets I'm Watson how I'm gonna take
your I've been very good so let's just
say we asked Watson how can you help me
with it so as you can speed answers I
can help you customize your pizza in
following ways sighs toppings crust
thickness and sauce now if you know from
what some of my previous videos
I'm actually not to make much of a thing
of sauce on my pizzas so I'm going to
say uh I don't really like sauce on that
pizza all right and it says I like no
sauce and it's set that perimeter all to
say you know what this user does not
want sauce then I'm going to say you
know what I'm very hungry give me the
biggest pizza
you have now again what's going to
happen is if I send this it will
understand that I want an extra-large
now again Watson is good great
just great a sort of filtering out the
noise which is I'm very hungry and so
you know it understands that hey that's
just conversation talk with humans enjoy
but it does it doesn't really understand
what that means it doesn't know emotions
and so give me the biggest pizza you
have biggest pizza it really focuses on
that and is even say you know what they
want extra large and since as you know
what you'll give them an extra lunch now
let's say I want to see what type of
class they have so what types of trust
do you have all right
I can ask and it says you know what you
can choose between thin and thick crust
so now I mean well usually I don't think
so Oh thin all right sounds good then
cross you know how toppings can I
actually have on this pizza okay so zip
you can see it says that I can have
three toppings maximum and I can choose
between cheese chicken all ups and
pineapple so now let's say I want cheese
all right but then again I didn't tell
it which topping I'd like that to be so
I'd say first topping is cheese all
right um second sprinkle some chicken
for example Oslo again it's good it's
Frost or fill 20 outfit noise sprinkle
some it removes that and said second
topping chicken okay so if I send that
as you can see alright so our second
topping this chicken I know if I say ah
third please add some cheese again as
you can see it filters out the noise and
it says all right so it's throat topping
is cheese and now let's just say we're
done I love just before we and the
conversation we want to ask how long
will it take to a lot and as you can see
it says the pizza should arrive within
30 minutes or it's free and I can say
already on
and as you can see it so sounds great
please proceed to the next page where
you can see final price and say your
shipping address and billing details and
just wats pizza does not get there in 30
minutes
it's free goodbye and that was a quick
demo of the IBM Watson conversation
service now what I'm going to do is I'm
actually going to put this into an
application iOS application includes
some sort of contextual analysis and
then I will be of course putting it into
another application now that's going to
be it for this tutorial now again today
we have created an application for users
to essentially be able to converse with
Watson in order to basically create this
sort of chat bot that users can converse
with in order to sort of feel like
they're talking to a human but they're
not actually talking to a human I know
it feels very natural it's not just like
you know what here's Pizza Pizza online
form where you fill in all these sorts
of details instead it's very natural hey
I don't want sauce i feets I'm not too
big of a fan of that I want thinner
crust but you can actually talk to it
for example you can incorporate speech
to text and text to speech again that's
going to be all for this tutorial thank
you very much I hope you enjoyed and if
you did please make sure to leave a like
and subscribe to the YouTube channel it
really does help out a lot and of course
even if you'd like to see my future
content and tutorials all right so now
again if you have any questions
suggestions feedback you can of course
leave down the comments email me
attachment gmail.com tweet to me at 10 G
Manny and that's going to be it thank
you very much again subscribe to the
channel turn you or your lock on content
you want some more of it
and that's gonna be it good bye