►
Description
This 16 minute video is part 1 of a 3 part series on building conversational bots for Microsoft Teams. This part introduces the Language Understanding Intelligent Service (LUIS) in Microsoft Azure, and shows how to use it to predict the user's intent and details about the user's request (called entities). Sample code in C# is provided; the concepts are the same in other languages.
Presenter: Bob German (Microsoft) | @Bob1German
Supporting materials:
• Sample bot code: https://aka.ms/ConsultingBot
• LUIS: https://luis.ai
• Microsoft Bot Framework: https://dev.botframework.com/
• Teams Dev Center: https://developer.microsoft.com/en-us/microsoft-teams
PnP videos for Teams developers: https://www.youtube.com/playlist?list=PLR9nK3mnD-OWG3R399jMSrt1eFa5KErvc
PnP videos for Teams low-code developers: https://www.youtube.com/playlist?list=PLR9nK3mnD-OUEuwES5Zn66qUdUdD60PCZ
A
Hi,
my
name
is
Bob
German
and
this
video
is
building
a
conversational
bot
for
Microsoft
teams,
part
1.
This
video
will
show
you
how
to
build
a
bot
that
takes
commands
using
natural
language,
interaction
with
teams.
Users
in
this
video
I'll
show
you
how
to
use
Lewis
the
language
understanding,
intelligent
service
in
Azure
to
predict
the
users
intent
and
any
related
information
about
what
they're
trying
to
do.
I'll
show
you
how
to
call
Louis
and
clean
up
the
entities
that
it
returns.
So,
let's
dive
in.
A
So,
let's
begin
by
considering
the
different
types
of
conversations
in
teams
inside
of
a
team's
channel.
There
are
many
users
involved
and
the
conversation
is
threaded,
so
it's
easy
to
create
a
new
conversation
accidently.
Instead
of
replying
to
a
thread,
you
also
have
to
add
mention
the
bot
with
each
interaction.
This
makes
it
hard
to
have
a
back
and
forth
dialogue
with
bots
for
group
conversations
and
channel
conversations,
I'm
going
to
suggest
a
different
approach
using
adaptive
cards.
A
The
approach
in
this
video
is
more
of
a
back
and
forth
conversation
and
is
more
suited
for
one-on-one
or
personal
apps.
So
imagine
I'm
a
consultant
and
I'm
just
finished
up
at
a
job
and
walking
out
to
my
car
and
I,
pull
out
my
Android
phone
and
bring
up
Microsoft
teams,
so
I
want
to
remember
and
want
to
build
my
hours
as
easily
as
possible,
so
I'm
going
to
use
the
voice
feature
of
the
phone
here
today.
A
I
worked
three
hours
on
the
fourth
coffee
project,
but
the
teams
based
bot
is
right
there
to
pick
up
my
mic
project
and
confirm
it
and
I'm
done
now.
Sometimes
users
don't
provide
all
the
details.
I
worked
on
a
project
which
project
while
the
bot
asks
contoso,
but
we're
gonna
find
out
there's
more
than
one
contoso
project.
First,
one
no
problem:
the
bots
gonna
lead
me
through
three
hours.
A
Okay,
now
it
has
all
the
information
and
once
again
it
can
go
ahead
and
build
a
client
for
the
work
so
to
make
this
work.
I
used,
Lois
a
language
understanding,
intelligence
service
in
Azure.
Luis
turns
utterances
that
is
whatever
the
user
said
in
too
intense
and
entities.
What
does
the
user
want
to
do
book
me?
A
flight
in
this
example
and
the
entities
Berlin
is
where
I'm
coming
going
from
to
Paris
on
Wednesday.
A
Those
are
the
entities
and
typically
the
the
intent
is
going
to
be
a
verb
and
the
entities
are
going
to
be
direct
or
indirect
objects
of
that
verb.
So
you
can
build
an
update.
A
model
in
Luis
and
I'll
show
you
that
in
just
a
moment,
and
then
query
that
model
from
your
bot
to
figure
out
what
your
user
wants
you
to
do
now
in
practice,
you're
not
going
to
get
perfect
results
back
from
Luis,
you're
gonna,
get
predicted,
intent
and
predicted
entities.
A
Your
bot
still
needs
to
handle
itself
appropriately
if
it
can't
get
all
the
information.
So,
what's
typical
is
that
you're
gonna
actually
send
the
message
text
the
utterance
off
to
Luis
and
if
you
don't
have
any
idea,
if
Luis
can't
even
predict
what
the
intent
is,
then
you're
gonna
get
a
generic
response
or
maybe
use
Q&A
maker
for
a
general
response.
A
If,
however,
you
know
what
the
intent
is
with
reasonable
assuredness,
then
you
can
go
in
and
try
to
get
the
entities
and
any
that
are
missing.
You
can
prompt
the
user
using
dialogues
so
in
this
video
I'm
going
to
show
you
Luis
how
to
create
the
model
and
how
to
query
the
model,
and
then
the
next
video
I'll
show
you
how
the
dialogues
work
so
that
we
can
refine
this
down
to
the
exact
information.
A
Slash
Consulting
button,
so
you're
gonna
want
to
start
by
navigating
to
Luis
ai
and
logging
in
with
an
account
that
has
a
juror
admin
permissions,
and
you
can
see
here,
I've
gone
in
and
selected,
my
adjure
subscription
and
what's
called
an
authoring
resource,
which
is
the
location
of
the
Louis
instance,
where
I'm
going
to
put
my
application
model.
So
you
now
create
this
right
inside
of
the
azure
portal
and
it
actually
gives
you
a
lot
more
control
and
notice
that
the
pricing
is
still
pretty
good
here.
A
A
Okay
and
now
I.
Don't
really
need
this
little
tutorial,
so
I'm
going
to
skip
by
that
and
go
right
to
the
build
tab
where
you
can
see
the
intents
and
entities
now
we
already
have
an
intent
here,
which
is
the
non
intent,
and
you
really
do
want
to
come
in
here
and
put
some
counter
examples
of
things
that
aren't
ordering
pizza
or
aren't
ordering
food
in
this
case.
Now,
obviously,
you
want
things
that
are
going
to
be
common
mistakes.
A
You,
okay,
so
a
few
different
ways
and
and
a
lot
of
work
can
go
into
the
end
of
this
and
figuring
out
how
people
of
different
cultures-
and
things
like
that
might
might
interact
with
your
bot.
So
now,
let's
go
in
and
set
up
the
entities,
so
pepperoni
is
a
flavor,
so
I'm
gonna
make
a
new
entity
called
flavor
and
I
could
make
it
a
simple
entity
which
is
just
a
string.
A
That's
taken
out
of
the
context
of
the
sentence
structure,
a
hierarchical
entity
which
would
be
something
like
a
geography
that
has
a
hierarchy
to
it.
Composite
is
where
multiple
words
are
coming
together
to
build
a
bigger
entity.
Phrase
I'm
just
going
to
use
a
list
entity
because
it's
kind
of
the
most
reliable-
and
this
is
a
flavor
and
I'm
just
going
to
leave
it
at
that
for
now
I
kind
of
prefer
to
edit
these
over
in
the
edit
E
and
in
the
entity
editor.
A
And
now,
let's
go
over
to
entities
and
fill
in
some
different
flavors
notice
that
it
was
already
smart
enough
to
suggest
some
Oh
bacon.
How
could
we
forget
bacon
mark
would
never
forgive
me
and
since
they
put
in
him
well,
we
got
to
put
in
pineapple
just
to
annoy
people
who
think
that
pineapple
never
belongs
in
a
pizza
and
now
I'll
go
over
and
do
my
sizes.
A
A
So
here
I'm
actually
asking
it
to
do
some
magic
with
the
date
last
Tuesday
and
also
to
translate
from
minutes
into
hours,
and
that
works
well.
Also
notice
that
I'm
going
to
be
able
to
handle
extremely
I
want
to
handle
extremely
terse
requests,
because
in
this
kind
of
an
application,
people
may
not
always
take
the
time
to
type
everything
out.
A
A
The
entities
are
kind
of
interesting
because
I
actually
have
two
date/time
entities.
My
date/time
entity
could
really
be
one
of
two
things.
It
could
be
the
day
that
I
worked
or
the
time
that
I
worked
yesterday.
I
worked
for
twenty
minutes.
Twenty
minutes
is
a
date
time
and
yesterday
is
a
daytime,
so
actually
created
two
different
roles
in
here,
and
if
you
look
carefully
at
my,
if
you
look
carefully
at
my
utterance
as
you
can
see,
that
I've
actually
tagged.
A
Not
only
is
this
a
date
time,
but
this
date
time
contextually
is
the
time
worked
versus
this
other
one,
which
is
contextual.
They
understood
to
be
the
day
that
I
worked.
So
if
I
test
this,
for
instance,
you
can
see
that
it
correctly
determined
that
the
date
time
today
was
the
day
that
I
worked
and
that
the
20
minutes
was
the
time
that
I
worked.
So
when
you're
done
with
your
app,
you
want
to
publish
it
and
once
you've
published
it,
you
can
actually
go
in
and
export
the
app.
A
You
also
need
to
copy
the
Loess
app
ID
and
app
key
into
the
app
settings.
Json
file
in
your
project
notice
that
I
can
export
and
import
versions
of
the
app
or
I
can
also
do
it
from
my
dashboard
here.
I
can
actually
I
mean
from
the
appleís
here.
I
can
actually
come
in
and
import
or
export
from
these.
A
So
it's
a
nice
practice
to
export
and
put
it
in
to
put
the
JSON
that
you
get
right
into
your
project
in
Visual
Studio,
and
so
here
you
can
see
inside
of
my
repo
I've
got
the
side
of
my
project.
I've
got
the
exact
Louis
ready
to
be
imported
and
used
along
with
my
project
now.
In
order
to
make
this
work,
I
actually
had
to
use
a
couple
of
new
get
packages.
A
A
A
What's
good
of
it.
What's
our
goal
right,
let's
take
a
step
back.
What
I
really
want
to
do
is
get
this
information
right
here,
because
I
really
need
to
know
what
is
it
that
I
really
need
to
know
enough
to
go
and
update
the
backend
system
right?
So
this
is
what
I
really
need
to
update
the
again
system
as
I'm
going
and
gathering
the
information
I
have
this
more
detailed
class
that
has
all
of
the
information
that
I've
built
up
so
far.
A
So
what's
gonna
happen
is
that
I'm
gonna
get
as
much
of
this
as
I
can
out
of
Louis,
but
it's
not
going
to
be
complete.
Then
my
bot
dialogue
is
gonna.
Look
through
see,
what's
missing
and
ask
for
all
the
missing
values.
So
basically,
everything
in
here
is
going
to
be
about
filling
in
this
data
structure
and
then,
when
it's
finally
done
I'll
be
able
to
fill
in
this
data
structure
that
I
pass
to
the
back
end.
A
So
this
recognizer
class.
What
it's
going
to
do
is
it's
going
to
actually
go
and
create
a
new
consulting
request.
Details
object
right!
That's
what
it's
going
to
give
us
back,
based
on
whatever
it
gets
from
the
utterance
that
the
user
typed,
which
is
inside
of
the
turn
context.
So
the
first
thing
I'm
going
to
do
is
connect
to
my
Lewis
application
and
get
a
new
Louis
recognizer
for
that
and
on
line
34
I'm
actually
going
to
make
the
call
to
Louis.
A
So
then,
I'm
gonna
call
recognizer
get
top-scoring
intent
and
it's
going
to
return
two
variables,
the
intent
and
the
score
now
I'm
going
to
start
trying
to
get
some
of
these
other
values,
so
the
person
name
which
is
used
in
the
other
intent.
Actually
the
project
name,
the
date/time
values
the
day
worked.
These
are
so
it
turns
out
that
not
a
hundred
percent
of
the
time
is
it
going
to
be
able
to
tell
whether
a
date
time
is
my
day
or
the
time
that
I
worked.
A
A
So
that's
going
to
be
the
next
video
I'll
show
you
how
to
build
a
dialogue
that
will
go
through
clean
up
all
the
data
and
get
it
to
the
point
where
you
can
actually
update
the
request
for
the
user
thanks
for
watching
this
Microsoft
365
patterns
and
practices.
Video,
if
you
like
this
video,
please
subscribe
to
the
patterns
and
practices
youtube
channel
at
AKA
ems,
slash,
SPP,
NP
of
videos,
I'm
Bob
Sherman,
you
can
follow
me
on
Twitter
at
Bob,
one
German
and
please
check
out
my
blog
at
Bob,
one
German
comm!