►
From YouTube: Getting Started with SharePoint Syntex
Description
In this 48-minute Developer - focused demo, Microsoft Program Manager Sean Squires introduces PnP community viewers to SharePoint Syntex – a product under the Project Cortex umbrella. Syntex automates content classification and management for better content discoverability. Specifically classifies unstructured content - document understanding and structured form processing using AI integrated into SharePoint. Train Syntex model with very few documents. Models are associated to libraries and lists. Syntex went GA in October, is considered an add-on and 30-day trials are available. This PnP Community demo is taken from the SharePoint (PnP) Community call – November 2020 – Monthly sync call recorded on November 10, 2020.
Demo Presenter: Sean Squires, Program Manager (Microsoft) | @iamseansquires
Supporting materials:
• Document - Create a form processing model | https://docs.microsoft.com/en-us/ai-builder/create-form-processing-model
• Announcement - Announcing SharePoint Syntex | https://aka.ms/SharePointSyntex/Announce
• Website - SharePoint Syntex | https://aka.ms/SharePointSyntex
• Documentation - Introduction to Microsoft SharePoint Syntex | https://aka.ms/SharePointSyntex/documentation
• Resources - Microsoft 365 Content Services/Partner Program | https://aka.ms/SharePointSyntex/partners
Learn more:
• Learn more about the Microsoft 365 Patterns and Practices community at: http://aka.ms/m365pnp
• Visit the Microsoft 365 Patterns and Practices community YouTube channel: https://aka.ms/m365pnp-videos
A
What
is
syntax?
Yes,
thank
you
for
the
time
and
happy
to
share
with
you
what
I've
been
spending
gosh
the
last
year
and
a
half
working
on
so
yeah
good
morning.
Everyone!
Thank
you.
Let's
do
this.
I'm
gonna
turn
off
video
just
because,
but
I
am
here
hello
and
let's
share
the
screen,
so
I'm
gonna
throw
a
bunch
of
information
at
you
guys
and
then
we're
gonna.
You
know
demo
as
much
as
we
can
in
the
time
that
we
have
together
this
morning,
but
just
wanted
to
take
some
time
to.
A
You
know,
share
with
you
what
syntax
is
what
happened
to
project
cortex
what's
going
on,
so
I
thought
it'd
be
helpful,
not
knowing
where
everybody
is
in
their
journey
learning
about
this
new
product
add-on.
To
just
give
you
some
framing
and
help
you
understand
where
this
sits
in
the
broader.
What
is
now
we
kind
of
refer
and
refer
to
internally
as
the
project
cortex
initiative,
because
the
evolution
of
it
really
has
been.
A
We
were
working
on
what
initially
was
going
to
be
just
one
larger
product
and
then
that's
gone
through
a
couple
of
iterations
and
evolutions
and,
while
there'll
be
a
lot
of
connection
between
the
various
pieces
that
the
larger
team
is
working
on,
they're
really
started
to
evolve
a
very
different
set
of
capabilities
that,
while
they
complemented
one
another,
we're
also
targeting
different
use
cases
and,
in
some
cases,
different
members
of
an
organization.
And
so
syntax
is
really
around.
A
You
know
to
make
sure
that
things
are
a
tagged
at
all,
but
also
consistently
tagged,
and
that
metadata
is
really
such
a
cornerstone
of
a
really
rock
solid
content
management
system,
because
it
can
really
help
in
so
many
ways.
You
know
a
lot
of
how
we
look
at
it.
I
always
refer
to
like
these
sort
of
three
solution
pillars.
You
know
you
have
good
metadata,
it's
going
to
help
in
your
search,
retrieval
it's
going
to
help
with
your
downstream
business
processes.
A
While
we
were
working
on
this
and
really
just
saw
a
lot
of
not
just
the
pain
points
that
folks,
like
you
were
reporting
to
us,
but
also
just
the
opportunity
to
really
optimize
savings.
Honestly,
you
know
move
people
to
higher
productivity
savings.
You
know,
get
productivity
savings
and
move
people
onto
higher
tasks,
because,
let's
admit
it,
none
of
us
like
to
manually
tag,
metadata
and
files
that
we're
hosting.
A
So
we
saw
opportunities
to
really
move
into
that
space,
and
so,
where
cortex
really
started
was
we
were
looking
at
sort
of
across
the
spectrum,
and
I
I
wanted
to
bring
up
this
slide.
Real
quick
just
to
kind
of
show
how
the
evo,
well
not
the
evolution
but
kind
of
where
we're
at
at
the
moment
is
project
cortex
kind
of
we
now
have
two
separate
products.
We
have
like
sort
of
the
knowledge
curation
discovery.
A
This
is
work
that
you've
heard
us
talk
about
at
conferences,
and
this
is
where
we
mine
your
existing
content
to
help
make
connections
and
identify
topics
and
connect
experts
with
the
content,
and
then
you
can
curate
that
and
that
really
helps
build
up
that
knowledge
base.
The
other
side
of
it
is
the
syntax
side
where
you
have
all
that
content.
A
How
can
we
help
you
with
ai
to
better
organize
that
information,
and
so
it's
a
little
bit
more
intentional,
a
little
bit
more
human
first
and
really
where
that
landed
was
just?
How
can
we
give
you
an
easy
way
to
build
ai,
app
capabilities
to
better
automate,
the
classification
of
your
content
and
the
extraction
of
metadata
from
that?
A
A
Our
ai
teams
have
just
been
building
some
very
innovative,
ai
capabilities
that
we
have
integrated
into
the
product,
and
I
just
point
this
out,
because
this
is
for
those
of
you
who
are
familiar
with
traditional
machine
learning
it
it's
it's
very
powerful,
but
it
takes
a
lot
of
time
takes
a
lot
of
training.
It's
typically
a
bit
opaque.
A
Is
it
kind
of
puts
the
human
first
and
at
the
center
of
the
equation
and
allows
you
to
work
with
a
smaller
sample
set,
do
a
limited
amount
of
labeling
and
then
review
those
results
more
real
time
and
in
in
effect,
kind
of
shorten
that
that
training
process
and
get
to
a
point
where
you
can
go
great?
I
can
now
very
quickly
and
easily
train
a
model
to
understand
that
this
is
a
contract,
and
this
is
not,
and
once
I
have
that
contract
then
do
some
additional
training
to
say.
A
I
want
you
to
identify
and
identify
certain
entities
from
that
contract
and
then
write
that
stuff
out
to
to
a
sharepoint
library
and
so
just
to
kind
of
put
this
in
frame,
because
you've
probably
also
heard
if
you've
attended,
listen
to
any
of
the
talks.
I
was
going
to
say,
attended.
We
have
all
been
in
quarantine
for
too
long,
so
listen
to
any
of
the
talks
that
we
have
given
in
the
last
year.
A
There's
we
kind
of
see
this
as
a
bit
of
a
spectrum
and
there's
different
capabilities
that
we
provide
in
this
release
of
syntax.
Obviously,
on
the
far
left,
you
have
just
sort
of
what
we
have
today.
You
know
the
manual
ability
you
know
to
leverage
metadata
and
tag
content
and
you
know
set
up
content
types
and
all
that,
and
I
will
just
use
that
opportunity
to
make
a
plug
that
as
a
part
of
this
effort.
A
As
many
of
you
know,
we
really
doubled
down
and
reinvesting
in
the
managed
metadata
service
and
a
lot
of
that
work
has
gone
out
already
and
is
available
to
you.
There
are
some
additional
enhancements.
We're
going
to
continue
to
build
on
the
management
of
data
service
and
invest
there,
and
some
of
those
investments
are
reflected
in
the
the
syntax
product,
but
there
will
be
also
additional.
You
know
improvements
made
to
the
service
itself,
moving
down
to
the
right
real
quickly.
A
You
know
just
more
text
analytics
where
imagine
just
being
able
to
upload
a
document
and
have
it
automatically
extract
key
phrases
using
the
azure
cognitive
service
apis,
and
so
that's
work.
That's
still
in
development,
but
certainly
a
really
neat
area.
When
you
don't
need
to
do
a
bunch
of
customization,
you
just
need
to
get
some
tags
out
to
maybe
improve
search-
or
at
least
just
to
you
know,
drive
some
general
processes
without
a
lot
of
customization.
A
But
if
you
have
specific
customizations
or
specific
entities
that
you
want
to
extract
from
your
documents,
then
it's
best
to
you
know:
go
into
the
custom
space
and
that's
really
where
syntax
shines
and
just
to
be
clear.
We
provide
two
different
technologies
here.
These
are
using
two
different
underlying
ai
apis.
A
And
what
we
do
is
we
do
all
the
heavy
lifting
for
you
in
terms
of
you
know,
as
you
train
the
model,
then
what
we're
going
to
do
is
take
that
model.
We're
going
to
you
know,
put
it
into
a
power
automate
flow
write
everything
back
to
the
library,
do
all
the
setup
for
you.
So
it's
a
pretty
seamless
experience
and
when
you're
done,
your
model
is
ready
to
go
the
document,
understanding
ones.
Those
are
built
on
some
different
technology
for
those
who
are
familiar
with
the
work
that
we've
been
doing
in
the
ai
space.
A
This
is
working
off
of
the
louis
apis,
and
what
this
is
doing
is
this
is
more
tightly
integrated
into
sharepoint,
and
this
allows
you
to
build
models
in
a
new
site.
Template
called
the
content
center
and
with
that
and
I'll
show
you
that
in
a
second
and
what
you
can
do
with
that
content
center
is
we
provide
a
training
experience
that
allows
you
to
effectively
build
these
models,
and
these
are
a
little
bit
different,
because
they're
being
centrally
created
and
managed
in
the
content
center.
A
The
difference
here
is
that
these
models
can
be
applied
to
multiple
libraries
so
or,
and
they
can
also
be
reused.
You
know
you
can
apply
multiple
ones
to
the
same
library,
so
they
have
a
little
bit
more
flexibility
right
now
in
their
usage
and
then
the
difference
between
the
two
is
how
what
kind
of
content
you're
needing
to
classify
and
extract
information
from,
and
we
we
recognize
that
this
is
a
bit
of
a
a
cognitive
load
and
you
might
say
well
sean.
A
So
you
know
office
files,
pdf
things
where
you
have
a
lot
of
just
text,
but
you
have
some
consistency
like
in
that
case:
you'll
see
that
service
agreement.
So
if
you've
got
like
a
contract,
that
has
a
lot
of
text
and
you
want
to
extract
information
from
that.
Like
maybe
the
you
know
the
start
date
or
what
the
you
know
who
the
vendor
is
or
what
the
terms
of
agreement
are,
you
can
be.
A
You
can
extract
those
clauses
because
they're
sort
of,
like
you
know
loosely
you
know,
found
in
the
paragraphs
themselves,
whereas
the
form
processing
you
know
using
the
form
recognizer
api
in
ai
builder,
that
one's
a
little
bit
more
structured
that
one's
better.
For
you
know
digital
forms
and
invoices
things
where
you
have
a
real,
clear,
more
structured
format
where
you
have
a
clear
key
value
pair
and
what
you
want
to
do
is
say
hey.
A
A
A
You
now
have
the
ability
to
highlight
undetected
fields,
you
know.
So
if
it
missed
that
date
or
that
invoice
total,
you
can
go
and
train
them
all
to
say:
hey
you
missed
this.
Let
me
highlight
it
for
you
and
walk
through
that
training
and
again
since
you're.
Only
you
know
training
it
on
a
half
a
dozen
files.
It's
it's
a
much
easier
labeling
and
training
exercise,
certainly
quicker,
and
then
once
you're
done.
A
A
I
just
want
to
give
you
guys
some
more
framing
just
so
you
understand
all
the
various
pieces
that
are
at
work
here
is.
We
also
do
some
work
to
those
libraries,
so
there's
a
bunch
of
things
that
happen
with
those
libraries.
When
you
apply
these
models
to
them,
we
want
to
make
sure
that
users
are
aware
that
hey
there
are
models
applied
to
this
library
and
at
runtime
we
give
you
a
notification
banner
that
lets.
You
know
that
some
additional
processing
is
happening.
A
We
also
have
integrated
panel
detail
experience,
so
you
can
kind
of
open
up
a
panel
and
see
what
model
or
models
are
applied
to
that
library.
And
then,
if
you
are
the
owner
of
those
models,
you
can
jump
directly
into
an
edit
experience
if
you
need
to
work
with
those
models
and
then
real
quickly,
we're
not
probably
not
going
to
talk
much
about
this
more
today,
but
I
will
call
out
that
first
one
around
taxonomy
extractors
there
are
some
additional
metadata
capabilities
that
you
know
with
project
cortex
at
large,
but
more
specifically
with
syntax.
A
Currently
is
the
ability
to
use
manage
metadata
fields
when
you're
training
extractors
for
your
document
understanding
models?
This
is
super
cool
I'll
show.
This
I'll
talk
a
little
bit
more
about
that
one
in
a
second
and
then
there's
some
additional
just
things
that
we
did
in
the
managed
metadata
service
itself
from
an
administrative
capability.
A
Finally,
as
a
first
step
in
solving
the
conundrum
of
our
content,
type
syndication
sucks
in
sharepoint
online,
we
hear
you
and
we
are
fixing
it.
The
first
step
is
with
content
type
push.
This
is
effectively
the
ability
to
take
an
enterprise
content
type
and
push
it
to
a
hub.
So
if
you
have
a
site
designated
as
a
hub,
then
what
we're
going
to
do
is
we're
going
to
push
that
content
type
to
that
hub
and
then
it's
going
to
be
automatically
applied
to
any
new
lists
or
libraries
in
the
associated
sites.
A
To
that
hub,
I'm
going
to
skip
over
this.
I
think
you
guys
for
those
who
are
joining
you're,
probably
already
you
know
on
board.
I
don't
need
to
convince
you
about
the
value
of
metadata
and
taxonomy
and
information
protection
labels,
but
clearly
the
benefit
that
we
see
here
is
really
just
again
automating.
A
All
of
this
really
gives
you
guys
that
flexibility
to
take
advantage
of
what
the
ai
app
models
can
do
in
terms
of
that
classification
generating
that
metadata
from
the
documents
so
that
you
can
then
just
really
start
to
you
know
improve
the
quality
and
consistency
of
your
search
results,
any
business
process,
automations
and,
of
course,
just
consistent
compliance
governing.
So
I'll
talk
a
little
bit
about
that
one
in
a
second.
A
I
don't
remember
if
I
have
a
demo
set
up
for
this
at
the
moment,
a
bunch
of
my
demos
in
the
last
week
or
two
have
been
sort
of
reset,
but
with
the
mip
sensitivity
and
retention
label.
Let
me
just
talk
about
that.
One
super
quick!
So,
as
many
of
you
know,
if
you
start
working
with
this
with
the
information
protection
labels,
you
can
define
those
things
and
define
the
classifiers
to
say:
hey.
If
I
have
a
certain
content,
type
go
and
stamp
it
with
a
particular
information
protection
label.
A
We
sort
of
are
accelerating
that.
If
you
will,
because
when
you
the
the
work
that
we
did,
there
was
currently
with
the
retention
label.
We
provide
the
retention
label
in
a
document
understanding
model
as
a
setting.
So
if
that
retention
label
is
defined
in
the
security
compliance
center
and
then
it's
published
at
the
content
center,
you
can
use
that
in
that
retention
label
apply
it
to
a
model,
and
so
when
that
model
gets
applied
to
a
library,
what
we
do
is
we
just
accelerate
that
application
or
that
stamping
of
that
retention
label.
A
So
what
we'll
do
is
at
run
time
when
that
model
runs
against
that
uploaded
file,
and
we
classify
it.
Then
we
also
will
stamp
it
with
that
associated
retention
label.
So
it's
not
to
displace
or
replace
what
the
you
know.
Information
protection
you
know
scanning
service
does,
but
it
can
do
it
a
lot
quicker
right.
You
can
do
it
at
a
time
of
upload
and
so
there's
that
benefit
of
getting
that
retention
label
stamped
on
that
file
a
little
bit
more
quickly.
The
sensitivity
label
worked
different
set
of
apis.
A
A
A
lot
of
this
stuff
has
been
shared
in
the
tech
community,
some
of
the
posts
that
seth
and
chris
and
others
have
shared,
and
I
think
even
some
stuff-
I've
shared
on
twitter,
but
I
just
wanted
to
sort
of
summarize
that
there's
a
bunch
of
pieces
that
are
a
part
of
the
syntax
release,
because
I
know
again,
you
guys
have
heard
like
wait,
cortex
was
being
shipped
and
then
it
was
called
syntax,
and
then
you
release
syntax
and
not
topics.
Yet
topics
are
coming
later
this
year
as
a
part
of
the
cortex
initiative.
A
So
right
now
the
syntax
work
it
just
went
live
last
month
there
are
trials
available
just
to
sort
of
head
off
a
question
that
has
come
up
part
of
the
syntax
licensing.
Is
you
get
service
credits
perform
processing
if
you
have
300
or
more
seats
and
that
initiate
that
essentially,
is
the
equivalent
of
getting
the
ai
builder
service
credit
capacity
add-on.
A
So
you'll
get
a
million
credits
each
month
and
if
you
the
trial
licenses,
because
by
default
they
only
have
25
seats,
just
be
aware
that
the
those
don't
come
with
credit.
So
if
you
don't
already
have
credits,
you
won't
be
able
to
try
out
the
form
processing
experience
with
the
trial
license.
We
are
looking
to
see
if
there's
some
ways
that
we
can.
A
A
lot
of
stuff
is
available
on
the
product
page,
including
some
fantastic
documentation
that
we've
put
together
a
lot
of
the
documentation
includes
embedded
videos
that
just
kind
of
highlight
some
of
the
experience
and
the
things
that
you
can
do
with
syntax
and
then,
of
course,
for
our
virtual
ignite
event.
Back
in
september,
we
did
a
bunch
of
videos
that
kind
of
walk
through
both
the
general
syntax
value
proposition
and
where
it
sits
in
the
broader
project
cortex
initiative.
But
then
we
also
did
a
bunch
of
drill.
A
Downs
myself
and
my
engineering
manager
even
walked
do
a
whole
session
where
we
walked
through
the
whole
document.
Understanding
that
model
building
experience
I'll
touch
on
some
of
that.
We
probably
won't
have
time
to
walk
through
all
of
that
in
the
remaining
15-20
minutes,
but
I
just
wanted
to
let
you
guys
know
that
there
are
a
lot
of
great
resources
out
there,
and
so,
if
you
haven't
checked
them
out,
please
do
feedback
is
always
always
welcome.
A
So
if
there's
something
that
you
think
is
incomplete
or
incomprehensible
or
missing,
or
it's
awesome
we'd
love
to
hear
all
of
that
feedback
so
that
we
can
make
sure
that
we're
getting
you
guys
the
information
you
need
to
take
advantage
of
this
new
product
add-on,
all
right,
I'm
going
to
pause
for
a
second,
because
that
now
I'm
going
to
just
jump
to
some
working
code
and
demo
stuff.
Is
there
any
questions
in
the
chat.
B
B
A
Yes,
I'm
sorry,
you
know
what
I
think
I
have.
I
put
a
licensing
slide
in
here
in
case
this
came
up:
yeah,
okay,
so
sharepoint
syntax
licensing,
there's
more
on
the
product
page.
By
the
way,
that's
gonna
do
a
much
better
job
of
explaining
this
than
I
do.
I
build
the
software.
I
don't
license
it
or
build
the
licensing
plan,
so
my
apologies.
If
I'm
trying
to
get
more
versed
on
some
of
the
nuances
here,
but
effectively,
think
of
sharepoint
syntax
as
an
add-on,
okay,
so
it
I
believe
that
there
was
talk.
A
Maybe
you
know
just
to
keep
ourselves
honest
here.
We
may
have
even
mentioned
that
ignite
2019,
that
this
would
probably
be
like
an
e3
e5
add-on,
I'm
sorry
included
in
e5
and
available
for
e3
or
whatever
it's
a
little
bit
more
broad
than
that.
Now
it's
just
a
general
add-on.
If
you
have
pretty
much
any
type
of
sharepoint
enterprise
license
or
commercial
license
or
education
license,
you
can
add
the
sharepoint
syntax
license.
You'll
see
it
in
the
add-on
section
of
your
microsoft.
A
You
know
admin
center,
you
know
where
you
can
purchase
products
and
things
like
that.
That's
also
where
the
trial
will
be.
The
plan
is
price
per
user.
It's
five
dollars
per
user
per
month.
If
you
have
300
or
more
seats,
then
you
also
get
included
in
that
in
that
purchase,
the
ai
builder
monthly
credits
of
one
million,
and
that
gets
it
doesn't
roll,
so
it
gets
reset
at
the
beginning
of
each
month.
B
A
That
might
change
in
the
future,
but
right
now,
no
the
depending
on
the
license,
the
the
what
the
license
is
going
to
do
is
it's
going
to
give
you
guys
the
ability
to
here.
Let
me
just
jump
back
up
to
this
slide:
real,
quick,
the
administration,
what
it
is,
what
it's
going
to
do
is
it's
going
to
light
up.
The
ability
to
configure
syntax
is
part
of
that
configuration
as
you
set
up
a
default
content
center
by
the
way
content
center
is
a
admin
site
template,
so
you
can
make
more.
A
A
It's
probably
gonna
evolve
into
something
like
a
modern
document
center
where
it
could
be
a
destination
portal
or
it
could
be
something
where
you
know
you
want
to
set
up
different
models
and
access
to
managing
those
models
by
department,
so
the
hr
content
center
versus
the
product
marketing
versus
the
operations,
so
you
can
create
additional
content
centers
through
the
admin,
sharepoint
site
creation,
experience
and
then
the
site
there's
also
I'll
just
mention
here.
You
have
the
ability
to
set
up
where
the
form
processing
option
shows
up.
A
So,
if
you
wanted
to
you,
could
it
can
be
off
for
everybody
on
for
everybody
or
you
can
restrict
it
to
specific
sites
almost
like
an
access
list?
So
then
any
library
in
that
site
they'll
get
the
affordance
on
that
library
ribbon
to
go
and
create
those
form
processing
models,
but
no,
we
are
not
the
main
filter
right
now
is
we
do
a
seat
count
to
determine
whether
or
not
you
get
the
additional
ai
builder
credits?
A
B
Cool
the
the
second
there's
so
many
questions
now
I'm
coming
we'll
we'll
follow
up
with
some
of
this
with
sean
after
the
call
as
well.
But
second,
a
big
first
set
of
questions
is
around
the
content
type
hub
and
the
bush
on
that
one
and
the
the
classic
questions
will
it
only
apply
to
new
lists
which
have
been
associated
on
the
sites
which
have
been
associated
to
a
hub,
or
will
it
actually
work
for
existing
lists
as
well.
A
So
currently
it
only
works
on
new
lists
and
libraries,
but
since
you
asked
I'm
going
to
ask
you
a
question,
do
you
want
it
to
be
retroactive
and
apply
to
any
existing
lists
and
libraries?
That
is
work
that
we
are
considering,
but
we
are
talking
to
customers
so
yeah.
If
you
can
chime
in
on
the
value
out
of
doing
that,
or
would
you
want
the
ability
to
control
how
that
works?
A
B
A
Okay,
so
I
want
to
just
fly
through
a
bunch
of
things
that
now
I
just
threw
a
bunch
of
words
at
you
and
some
pretty
powerpoint
slides
and
let's
just
go
see
what
some
of
this
stuff
starts
to
look
like
and
yeah
I'll.
You
know
what
let
me
see
if
I
can,
since
I'm
out
of
powerpoint
me
sorry
one
second,
let
me
just
get
out
of
that.
So
then
I
can
see
questions
in
my
other
screen.
A
Or
a
trial,
oh
the
trial,
so
so
the
trial
you
should
in
your
existing
tenant.
You
should
be
able
to
go
into
your
microsoft
admin
center,
where
you
can
see
all
your
license
products,
and
I
believe
you,
you
not
believe
you
have
the
ability
to
purchase
other
products
there.
There
is
look
for
syntax,
it
should
be
rolled
out
and
you
should
be
able
to
either
purchase
seats
or
select
a
trial,
and
the
trial
is
going
to
be
a
25
seat
trial
for
30
days,
cool,
okay,
so
real
quickly.
A
You
guys
have
probably
seen
some
of
this,
but
I'm
just
going
to
touch
on
some
of
the
things
that
I
talked
about
and
just
sort
of
help.
You
connect
the
dots
because
I
think
one
of
the
things
I
was
thinking
about
when
I
was
talking
with
vasa
about
you
know,
sharing
with
you
guys
kind
of
like
an
overview
of
syntax.
A
I
want
to
give
you
a
lot
of
framing,
because
I,
in
my
conversations
with
the
community
and
threads,
that
I'm
seeing
they're
still
like
people,
are
trying
to
understand
how
to
think
about
all
these
pieces
and
how
they
work
and
how
they
come
together
and
fully
acknowledge
that
you
know
there's
some
areas
where
we
want
to
optimize
this
and
we're
looking
for
parity
of
experience
across
these
things.
And
you
know
it's
a
bit.
You
know
I
I'll
be
fully
transparent
with
you
guys.
A
We
do
put
a
bit
that
cognitive
load
on
you,
where
you
know
you
have
to
think
about
wait,
a
minute
which
model
type
is
the
best
for
me,
and
so
let
me
just
kind
of
reiterate
that
one
again,
because
I
think
it's
also
helpful
to
talk
about
how
do
you
think
about
syntax?
You
know
we.
We
have
a
tab
program
that
we've
been
working
with
some
customers
and
one
of
the
things
that
has
been
helpful
is
to
help
people
understand
that
syntax.
Doesn't
you
know
solve
the
entire
business
process?
A
What
it's
intended
to
do
is
help
solve
pieces
of
it
that
then
accelerate
your
productivity
or
allow
you
to
you
know
optimize
those
business
processes
so
that
you
know
folks
can
do
other
things,
and
so
one
way
that
has
been
to
think
about
it
is
hey.
If
syntax
can
help
me
do
those
things
like
classifying
content
consistently
in
my
corpus
or
extracting
specific
metadata
that
can
then
drive
business
processes
awesome.
A
So
then,
what
you
want
to
do
is
you
want
to
think
about
what
are
those
business
files
in
my
corpus
that
need
that
type
of
attention
that
need
that
type
of
consistency
that
I
need
to
get
that
entity
data
from
metadata
from
so
that
I
can
drive
those
business
processes
or
ensure
that
they
get
properly
classified
to
comply
with
my
governance
file
plan,
and
so
thinking
about
that
strategically?
As
you
go
into
this,
I
think
helps
put
what
syntax
can
do
in
perspective,
and
then
you
can
think
about.
Okay.
A
A
You
can
start
to
see
like
what
models
are
being
applied
where
and
not
only
that
you
know
given
what
library,
whether
applied
to
how
many
files
are
being
processed,
because,
obviously,
if
you
want
to
make
this
investment,
you
want
to
make
sure
that
a
it's
being
used
and
b,
it's
doing
what
you
need
it
to
do.
You
know
and
really
helping
automate
those
processes
so
that
your
folks
can
do
the
more
interesting
work.
A
A
Okay,
I
am
not
I'll
I'll
find
that
window,
but
when
you're
in
the
model,
this
benefit
change
knows
what
this
is
is
this
is
a
sample
model.
So
when
you
create
a
new
content
center,
you
have
the
you're
going
to
see
an
affordance
to
import
this
model
and
what
it
does
is
it
activates.
This
benefit
change,
notice,
model
and
what's
helpful
about.
A
This
is
what
one
of
the
things
we
want
to
do
is
provide
folks
with
lots
of
opportunities
to
understand
and
learn
how
to
build
these
models
and
how
to
train
them,
and
so
what
this
does
is
when
you
instantiate
this
model
in
your
content
center,
it's
a
fully
functioning
model,
so
it
brings
in
all
of
the
training
files.
It
also
brings
in
a
bunch
of
entity
extractors.
A
Those
are
the
things
that
are
going
to
pull
out
specific
information
from
the
documents
and
write
those
to
the
corresponding
sharepoint
library
columns,
and
so
what
you
can
do
is
you
can
use
this
model
as
a
guide
to
understand
what's
happening,
and
so
you
can
also
interact
with
it.
So
you
can
see
like
how
was
the
classifier
trained
and
the
classifier
training
experience,
whether
you're
training
the
model
to
understand
how
to
classify
a
benefit
change
notice
from
one
that
isn't
as
well
as
from
an
extractor.
It's
a
pretty
similar
experience.
A
The
difference
is
that
you're,
looking
at
the
whole
document,
rather
than
something
in
the
document,
so
notice
that
the
first
step
is
to
label,
and
so
I've
brought
in
a
bunch
of
sample
files
here
and
I'm
just
going
to
go
ahead
and
label.
Yes,
these
are
examples.
This
is
an
example
of
a
benefit
change
document
and
then,
if
you
have
one
that
isn't
you're
going
to
label
it
as
not,
and
then
that
negative
example
is
super
important,
you
need
one,
and
this
is
important
because
it
helps
the
model.
A
This
is
where
the
you,
the
human,
come
in
to
play
when
you're
training
the
model-
and
this
is
also
what
kind
of
accelerates
and
facilitates
the
building
of
these
models
is,
you
know
more
than
the
model
does,
and
one
of
the
things
you
know
is
you
know
what
a
contract
is,
and
you
know
what
a
contract
isn't
and
so
by
giving
a
negative
example,
it
really
helps
the
model
to
very
quickly
understand
how
to
identify.
In
this
case,
these
benefit
change
notices.
A
I
want
to
jump
back
out
to
an
extractor
just
so
you
can
see
what
the
extractors
look
like,
because
they're
very
similar
in
experience,
in
this
case
these
extractors,
what
they're
doing
is
you're.
Similarly,
let's
go
get
the
date
of
change.
How
about
that?
So
with
the
data
change,
it's
very
similar,
you're
labeling
the
document,
but
instead
of
just
doing
a
positive
and
negative
now
what
you're
doing
is
you're
actually
highlighting
the
thing
that
you
want
to
extract
from
this
document.
A
So
in
this
case,
there's
a
change
note
date
in
this
document,
and
so
I
want
to
be
able
to
identify
it
and
you'll
see
here.
I
can
go
ahead
and
remove
that.
But
what
you
do
is
you
just
highlight
it,
and
then
that
saves
that
label
and
then
you
can
move
on
to
the
next
document
and
then
once
you've
labeled
a
sufficient
number
of
documents.
You
move
on
to
the
training
experience
and
what's
important
about
the
training
experience.
A
So
it's
almost
like
being
very
explicit.
The
other
type
of
one
that's
super
helpful,
is
the
pattern
explanation
and
the
pattern
explanation
is
where
you're
also
telling
that
the
model
training
it
to
understand
what
type
of
entity
this
is
is,
and
this
is
particularly
useful
for
things
like
if
you've
got
an
address
or
a
phone
number
or
a
url
or
even
in
this
case
a
date
so
you'll
see
here
that
if
I
look
at
this
date,
one,
what
we're
doing
is
we're.
A
We've
defined
this
as
a
pattern
and
you'll
see
that
what
we're
doing
is
we're
identifying
all
the
variations
of
how
a
date
might
appear,
and
this
helps
the
model
understand
how
to
identify
this
data
change
that
you
want
to
extract.
I
also
want
to
show
you
something
else:
that's
very
cool
about
this
to
facilitate
there,
because
you're,
probably
going
man
that
looks,
complicated,
sean
well,
we're
trying
to
make
it
easier
for
you
and
a
couple
ways
that
we
do.
That
is
one
we've.
A
Given
you
a
bunch
of
pre-built
templates,
these
pattern
and
in
some
cases,
even
some
phrase
list
patterns
and
I'm
sorry,
explanation
types,
and
so
you
see
these
templates.
What
they
do
is
they
already
have
all
the
patterns
laid
out
for
you,
and
so
what's
nice
is,
you
can
go
ahead
and
add
one
of
these
and
then
you
can
go
and
manipulate
it,
and
even
after
you
manipulate
it
and
go
ahead
and
apply
it.
A
Let
me
go
ahead
and
actually
save
this
just
so
you
can
see
that
it's
there
and
then
what
I
have
the
ability
to
do
is.
I
can
also
save
this
as
my
own
template.
So
what
we're
giving
you
the
ability
to
do
is
you
can
make
your
own
explanations
from
blank.
You
can
make
them
from
one
of
our
pre-built
ones
to
kind
of
bootstrap,
and
then
you
can
even
save
them
and
once
they're
saved
those
custom
ones
are
going
to
show
up
on
that
list
for
other
other
model
creators.
A
So
you
can
really
start
to
take
advantage
of.
You
know
you
reusing
these
explanations
that
are
unique
to
your
business
files
to
again
accelerate
the
creation
and
training
of
those
models,
and
then
once
the
the
once
you
do
that
there's
one
other
thing.
I'm
sorry
I
wanted
to
point
out
here
is
when
you're
also
building
training
these
fo
these
models.
Another
option
is:
let's
go
to.
A
A
A
Sorry,
so
you'll
notice
here
that
what
I'm
doing
is
I'm
trying
to
get
the
company
name
out
of
this
contract,
and
I
can
actually
build
an
explanation
that
helps
the
model
understand
where
to
find
the
company
name,
and
these
are
very
powerful
explanations,
because
they're
helping
the
model
understand
context,
and
you
typically
only
need
like
one
or
two
of
these
to
again
just
help
the
model
understand.
How
do
I
find
the
thing
that
you
want
me
to
extract,
identify
and
extract?
A
A
A
A
A
Well,
that's
really
stupid,
because
what
if
it
takes
a
little
bit
longer
than
your
model,
then
your
flow's
never
going
to
run
so
what
we
did
was
we
built
a
new
action?
We
do
a
couple
things.
We
built
a
new
action,
I'm
sorry
we
when
the
model
is
classified.
What
we
do
is
we
stamp
the
file
in
the
library
to
let
you
know
when
it's
been
classified,
and
then
we
have
a
new
action
in
power
automate
that
allows
you
to
build
a
flow
off
of
that
classification
date.
A
You
know
value,
and
so
what
you
can
do
is
you
can
wait
until
the
model
is
done,
processing
and
then
kick
off
the
rest
of
your
downstream
flow.
Okay,
I'm
looking
at
the
time.
I
want
to
tell
you
guys
one
other
thing:
real
quick
before
I
forget
it,
this
contract
services,
one
really
important,
because
I
think
it's
I
wanted
to
stress
again
that
a
lot
of
the
work
that
we're
doing
with
syntax
is
also
associated
with
the
investments
that
we're
making
in
the
managed
metadata
service
and
this
contract
services
field.
A
For
example,
I've
changed
this
model
a
few
times,
but
on
these
two
particular
records,
this
is
an
actual
managed
metadata
field.
It
is
here
it's
the
contract
services
and
it's
got
a
bunch
of
values
here
and
so
what's
very
cool
about
this.
Is
you
can
actually
in
this
model,
you
can
train
an
extractor
using
a
managed
metadata
field,
and
then
what
will
happen
is
when
the
document,
in
this
case,
looked
at
isaiah
langer
notice
that
the
contract
service
value
is
design.
A
Design
is
one
of
the
preferred
labels
in
this
term
set,
and
if
you
go
to
the
document,
the
actual
value
that
the
model
extracted
was
creative.
Creative
is
a
synonym
of
design.
So
what
we'll
do
is
when
you
run
the
model
against
this
file,
it's
going
to
extract
a
value
and
then
we're
going
to
match
it
to
the
associated
term
set
and
if
it
matches
either
the
preferred
label
or
the
synonym
we're
going
to
write
the
preferred
label
into
the
column.
A
So
it's
really
giving
you
that
ability
to
not
only
leverage
the
automation
that
the
model
provides
but
ensure
consistency
by
mapping
it
to
a
term
set
okay,
one
thing
we
did
not
get
to
touch
on
today.
I
knew
we
might
run
short
on
time,
but
I
did
just
want
to
highlight
for
you
guys
that
with
the
form
processing
stuff
it
does,
it
works
similarly
and
there's
some
interesting
work.
That's
also
happening
with
the
form
processing
that
that
area
is
continuing
to
evolve
and
syntax
users
will
get
the
benefit
of
those
improvements.
A
So
in
this
case,
you'll
see
that
there's
a
new
model
compose
feature
or
a
collection
feature,
and
you
can
read
about
it.
If
the
document
the
article's
been
updated,
it's
a
group
documents
by
collections
and
what's
powerful
about
this,
is
imagine
if
you're
getting
different
receipts.
I
know
sorry
different
invoices
from
different
different
shipping
companies
as
your
organization
and
instead
of
having
to
build
two
models.
A
If
you
know
that
you
just
want
to
build,
pull
the
invoice
id-
and
you
know
terms
of
laden
from
the
document
what
you
can
do
is
you
can
train
a
single
model
against
these
different
types
of
invoices
and
extract
the
same
metadata.
And
then,
when
you
do
that,
you
can
apply
that
same
model
back
to
the
library
so
notice.
I've
got
here,
two
different
types
of
invoices,
pr
contoso
po
and
then
these
optical
invoices.
A
They
both
are
laid
out
differently,
but
I
needed
the
same
information
from
them,
and
so
I
was
able
to
essentially
group
these
different
training
sets
into
different
collections
and
then
train
a
single
model.
So
it
gives
you
a
little
bit
more
flexibility
with
extending
the
power
of
a
form,
processing
model.
A
Okay,
again,
lots
of
information.
I
knew
this
was
going
to
happen.
We
were
going
to
run
out
of
time,
but
I
did
want
to
make
sure
that
you
guys
know
that
we
are
here.
We
want
you
to.
You,
know,
get
familiar
with
this
stuff.
Ask
questions
share
with
us.
Your
use
cases
there's
a
lot
of
information
out
on
the
product
page
and
also
in
the
tech
community,
and
also
on
the
virtual
hub.
So
please
check
that
material
out
and
let
us
know
if
you
guys
have
any
questions.
A
Yeah,
so
the
multilingual,
so
right
now
with
form
processing
form
processing,
only
supports
english.
You
have
a
little
bit
more
flexibility
with
with
the
document
understanding
models
that
tech
is
still
technically
only
supporting
english
at
this
time,
but
there
is
a
benefit
in
the
sense
that
the
way
that
the
model
does
the
processing-
it's
not
really
looking
at
words
so
much
as
it's
looking
at
tokens.
So
it's
looking
at
characters
and
arrangements
of
those
characters.
A
So
some
of
the
preview
customers
that
we've
worked
with
as
long
as
you
are
training
files
on
latin
alphabet
languages,
so
things
like
french
and
italian,
spanish,
german,
except
you
might
run
into
some
issues
with
diacritics
and
things
like
that
they
should
work.
But
official
support
is
just
english
only
at
this
time,
with
a
plan
to
support
more
languages
in
the
coming
year,.
B
Excellent,
thank
you.
Thank
you
clarifying
that
one.
We
got
that
one
in
a
video
as
well
and
and
we
do
apologize,
we
didn't
have
a
time
to
answer
all
of
the
questions,
but
we're
trying
to
follow
up
on
many
of
them
in
the
in
the
video
notes
or
the
blog
post,
which
is
about
this
community
call.
So,
but
thank
you,
sean
on
that.