►
From YouTube: Node.js Internationalization (i18n) Working Group
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
B
B
A
So
yeah
seems
like
nobody
joined
to
us
and
I
think
we
can
start
now.
So
what
zip?
Okay,
let
me
just
share
my
screen,
and
so
we
have
an
agenda
and,
as
you
see
here,
we're
starting
this
goal
and
yeah
the
first
point
in
our
agenda
is
machine
translation
risks,
and
so,
if
you
can
share
your
screen
and
and
make
your
game
stop
sharing
my
screen
and
go
ahead.
B
A
B
B
You
know
at
this
point
on
a
segment
level.
0
to
you
know
one
so
to
speak
or
0
to
learn
or
and
risk
yet
based
on
deep
learning.
It
supports.
Let's
call
it
all
languages,
definitely
all
the
languages
you
need.
It
works
for
machine,
translation
and
human
translation.
It
just
doesn't
even
know.
Is
it
a
machine
translation
or
is
it
a
human
translation
and.
A
B
The
API
and
console
are
the
main
ways
to
interact
with
it.
In
your
case,
probably,
the
API
is
more
interesting
and
we
happen
to
be
built
with
nodejs,
which
is
pretty
rare
for
a
chain
learning
pipeline.
We
haven't,
you
know,
machine
learning,
Python
under
the
hood
append
and,
of
course,
as
you
know,
we're
looking
to
support
some
open
projects.
B
We're
localizing
openly
are
doing
research
and
so
on
and
API
Docs
are
here
so
for
Mozilla
right,
which
is
also
an
openly
localized
project
and
the
open
source
project.
We
have
this
already
and
it's
a
shareable
and
linked
in-depth
in
those
notes.
So
you
can
see
here.
This
is
just
running
Mozilla.
Translations
Mozilla
does
use
Google,
Translate
and
then
post
edits.
A
B
Mozilla,
maybe
is
a
good
sort
of
example,
or
something
you
know
for
nodejs
I.
Think
right,
because
it's
sort
of
it's
an
open
project
that
maybe
more
languages
already
using
machine
translation,
not
using
custom
machine
translation,
and
so
not
that
advanced.
But
it's
still
pretty
advanced-
and
you
see
here
if
you
use
machine
translation,
that
with
Mozilla
that
most-
and
this
is
just
standard-
Google,
not
even
custom.
Most
of
them
are
pretty
low
risk,
the
vast
majority.
B
So
it's
actually
pretty
candidate,
especially
with
some,
and
it
doesn't
mean
everything
right
and
so,
whenever
we
you
know
you,
so
each
of
these
is
getting
a
risk
score
and
whenever
you
download
right,
these
are
they're
about
sixty
thousand
strings.
You
download
it
all
and
I've
also
linked
that
here
right.
So
you
you
go.
B
Put
the
link
to
that
I'll
add
that
afterwards,
but
so
what
you
see,
of
course,
this
is
just
the
machine
translations,
not
their
actual
human
translations.
You
see
a
few
catastrophes
I've
sorted
by
risk
right,
so
this
is
just
a
catastrophe
right
cookies.
Cookies
is
getting
translated
like
this
not
correct
in
a
technical
context.
Firefox
is
getting
translated
to
Firefox
the
turkey,
the
country
to
Turkey,
be
the
bird
and
so
on.
B
Okay,
so
this
is
not
a
good
ear
on
getting
translated
like
I
gurantee
like
Iran,
okay,
this
is
not
good,
but
if
we
go
look
at
the
big
picture
here
again
right
I
mean
I,
barely
scroll
down
and
already
we're
at
okay
18%
risk.
You
know
scroll
down
a
little
more
we're
at
3%
right,
so
the
vast
majority,
all
the
rest,
all
the
way
down
is.
It
was
translated
to
fine
with
machine
translations
if
you
believe
model
front.
B
So
you
know
it's
very
tempting
in
such
a
situation,
not
even
review
these
depending
on
how
good
we're
doing
you
know
to
skip
those,
or
at
least
use
them
as
a
placeholder
sort
of
the
way
you
used
crowd-sourced
translations
that
are
not
yet
approved
or
the
way
you
might
use
English.
So
so
that's
for
Mozilla
I've
done
something
similar
for
nodejs.
B
So
just
taking
the
English
sources
from
the
Spanish
is
the
largest
set
of
English
strings
that
you
have
about
fifty
thousand,
and
here
the
distribution
does
not
look
as
good,
but
let
me
compare
here
right.
This
is
using
a
custom
Mozilla
model,
whereas
this
is
just
using
a
generic
model
and
our
generic
model
may
be
harsh
on
some
of
your
translations
might
see
them
as
risky
right.
So,
for
this
is
a
great
example
right
so
actually
complement
those
I
looked.
B
Is
your
preferred
translation
for
add-ons
add-ons
with
a
capital
A
without
a
dash
is
a
kind
of
weird
word,
which
is
why
it's
considered
risky,
but
that
is
actually
what
you
want.
So
this
is
a
perfect
example
we're
having
the
custom
to
say
no,
no.
This
is
in
the
node.js
context.
This
is
how
we
want
to
translate
it
that
will
kind
of
avoid
that
false
positive,
save
your
effort
right
so
I
actually
went
through.
This
is
I
just
downloaded
that
whole
file
and
sorted
by
risk
here
so
I
did.
B
A
B
B
A
B
The
thing
too,
with
machine
translation
sometimes,
is
that,
although
the
source
side
is
a
factor,
you
know,
the
difficulty
does
increase
the
risk.
It
is
amazing
how
compared
to
human
translation,
how
it
can
do
very
well
on
such
hard
things
and
then
something
like
this
you'd
think
they'd
have
a
rules-based
file
or
something
like
just
to
avoid
this
kind
of
mistake.
B
Here
you
see
this
kind
of
stuff
right
so,
and
this
is
the
case
all
of
these,
where
I
would
suggest
not
simply
using
model
front
but
using
custom
machine
translation.
If
you're
going
to
miss
you
to
do
machine,
translation
or
I,
guess
crowded,
has
a
translation
memory
right
so
just
to
sort
of
read
it's
great
if
we
catch
it
for
you,
but
it's
much
better.
If
it
never
happens
right
because
at
the.
B
Again,
it's
a
small
fraction,
but
if
we're
talking
about
a
lot
of,
if
we
catch
a
lot
of
strings
for
you,
then
you
you
still
have
to
go,
fix
those
ones
or
they
don't
get
pushed
or
whatever.
And
so
you
want
to
reduce
that
so
overall
I'm
off
to
optimistic
about
this,
because
this
is
basically
easy
stuff
out
here,
we
have
again
Windows
right,
I
mean
last
all
of
this
stuff,
where
these
isn't
the
99%
risk
category.
If
we
go
back,
I
just
find
it
easier
and
to
look
at
it
in
model
front.
B
B
You
know
a
few
thousand
that
are
bad,
so
yeah
you're
close
right
and
you
get
a
little
bit
better
translation
quality
start
using
your
TM.
Add
some
do
not
translate
strings
and
you're
going
to
be
in
in
much
better
shape.
So
yeah
I
went
totally
off
the
script
here,
but
yeah
I
show
you
how
this
is
working
for
Mozilla.
The
links
are
here.
I'll.
B
Add
the
link
to
the
spreadsheet
as
well
now
the
challenges
I
see
for
nodejs,
basically,
is
you
want
to,
of
course,
save
on
your
human
translator
effort,
especially
if
you
really
want
to
have
a
hundred
hundred
languages,
or
something
like
that?
You
do
have
all
of
these
code
sources
right,
I
mean
just
tons
and
tons
of
stuff
like
this
right,
which
is
just
not
meant
to
be
translated.
B
It
depends,
and
it
depends,
is
the
hardest
thing
and
right
now
we
have
also
some
of
these
so
I.
Just
looked
out
of
curiosity,
because
that
this
Google's
gave
this
translation
and
I
said.
Okay,
how
do
you?
How
is
that
called
in
Spanish
or
what
is
no
js',
calling
it
right
now
and
it
turns
out
you
call
it
different
things
in
different
places,
and
so
that's
also
something
it's
like
cleaning
up
once
and
then
trained
correctly,
and
it
will
get
enforced
so
I.
A
B
Here
we're
talking
about
a
hundred
thousand
total,
which
is
not
a
lot
for
training
machine
translation.
Luckily,
our
model
is
multilingual,
so
collectively
this
100k
is
kind
of
what
we
think
about.
But
if
you
talk
about
training,
custom,
machine,
translation,
Google
or
Microsoft
generally
require
like
2
K
to
10
K,
minimum
or
recommended.
B
So
from
my
perspective,
you
won't
get
great
machine
translation
without
except
for
these
very
top
ones,
and
even
then
it
won't
be
great,
but
a
lot
of
the
stuff,
because
it's
technical,
it
can
benefit
from
dictionaries.
So
you
don't
really
need
the
machine
sort
of
the
very
fancy
machine,
translation,
sequence,
a
sequence,
kind
of
behavior.
You
need,
more
just
to
say:
hey,
never,
translate
Windows.
A
B
A
B
B
B
Yeah,
so
what
I
am
so
I
think
I'm
pretty
optimistic
about
what
you
can
do
in
this
case.
All
of
these
machine
translation,
API
zarnow
support
that
your
volume
is
not
very
high.
So,
however,
we
you
know,
do
the
machine
translation,
it's
not.
It
would
probably
fall
in
the
free
tier,
for
example,
on
Microsoft,
but
with
a
custom.
It
would
cost
something,
but
I
don't
know.
I
really
think
from
your
volume
could
be
something
like
two
dollars
a
month
now,
as
far
as
what
model
fronts
plans
here,
since
we
have
your
TM.
B
A
B
Like
we
do
also
for
Mozilla
eyewear
here
we
look
at
the
human
translations
and
there's
a
few
things
that
are
not
ideal
like
they
usually
translate
avatar
to
Avatar.
This
one
is
actually
legitimate,
but
in
context
but
not
not
ideal,
and
so
on
this
kind
of
thing
somebody
has
to
go,
look
at
it
right
so
same
deal.
Obviously
you
have
far
fewer
strings.
So,
if
you
in
their
case
they
need
to
look
at
a
thousand.
In
your
case,
you
might
only
need
to
look
at
100.
Oh
that
sounds
doable.
Then.
B
So
so
we'll
do
that
clean
out
any
bad
ones
that
we
don't
want
in
training
data
boost
the
accuracy.
I.
Don't
expect
this
model
with
this
few
training
segments
to
be
the
best
that
it
can
be
so
I
will
boost
it,
pulling
in
other
data
and
basically
increase
the
accuracy
and
also
try
to
cover
all
the
languages
that
you
don't
have
yet
so
we
can
get
to
something
like
so
have
pretty,
even
if
you're,
not
a
local
izing
them
yet
already
risk
prediction
customized
for
no
GS
working
well
for
100
languages
and.
A
B
Yeah
and
then
providing
that
you'll
just
have
your
API
key.
You
can
check
it
out
in
the
console
for
all
your
machine
translations,
crowd-sourced
translation
or
the
approved
translations
also,
so
the
human
translations.
You
can
also
send
us
I,
don't
know
what
what
you
want
to
do
right
if
you
want
to
make.
B
Check
those
again
right
at
the
end,
maybe
with
a
different
risk
threshold,
so
all
be
the
model
for
an
API
I.
Don't
expect
your
volume
to
be
very
high
I
think
we
discussed
that
before
and
then
optionally.
We
could
include
machine
translations
from
Microsoft
or
Google
in
those
you
can
get
them
via
our
API
or
you
can
call
them
separately
and
free
sorry
for
no
js',
of
course,
as
an
open
project
that
we
also
benefit
from
greatly
so
out
of
order
here.
But
that's
that
and
yeah.
There
are
some
open
questions.
B
A
We
have
projects
our
first
one
is
a
current
OGS
website.
Second,
one
is
I,
knew
no
js'
website
that
we
are
currently
working
on
it
and
third
one
is
API
Docs,
so
translation
for
API
Docs
is
never
been
life
and
we
are
like
just
started
investigating
in
this.
How
we
can
deliver
this
content
to
the
website,
because
currently
it's
showing
only
original
source,
English
and
knowing
the
fallback
know
any
translations
just
English
entity,
and
we
are
going
to
solve
this.
A
B
B
Okay
and
then,
of
course,
there's
the
question
I
think
it's
a
longer
discussion
within
nodejs
about
the
how
address,
if
you
want
to
go
write,
do
you
want
to
start
off
with?
This
is
a
suggestion
that
makes
it
much
faster
for
volunteers
to
approve
arm
and
much
faster
for
reviewers
to
review
because
of
the
getting.
A
B
B
A
This
like
this,
how
I
can
see
this
process,
but
at
the
same
time,
it's
like
a
question
for
model
front,
because
I
believe
that
guys
from
the
team
is
very
development,
developer
oriented
guys
and
they
are
not
want
to
have
deal
with
like
some
manual
parenting,
quiz,
JSON
file
or
excel
file
with
risks,
and
they
are
will
start
looking
for
some
way.
How
we
can
make
this
process
automated,
and
it's
like
question
for
you-
how
we
can
make
this
process
automated,
because
so
we
have
chromium
and
all
this
crowd
in.
A
We
have,
for
example,
we
start
using
machine
translations
how
we
can
say
to
crowd
in
okay,
how
we
can
take
from
your
site
the
list
of
strings,
that
we
are,
that
we
ready
to
translate
by
a
machine
learning
and
how
we
can
filter
it
on
crowd
in
silence.
Like
general
questions,
probably
you're
already
know
the
answers,
and
it
was
very
helpful.
A
B
B
B
But
I
guess
there
are
this
push
or
pull
right,
yeah
and
I
wouldn't
know
too
well
at
what
stage
you
get
a
string.
You
know
at
what
stage
does
a
string
go
from
your
repo
in
to
crowd
in
and
at
what
stage
does
a
string
come
from
crowding
back
and
then
do
you
have
some
kind
of
I?
Don't
know
RSS
or
some
you
know
pub/sub
at
all,
from
crowding.
B
A
Crowd
in
years,
Rajan
has
some
github
integration
that
we
are
using
so
like.
We
are
just
connected
our
target
repos
and
the
crowd
didn't
automatically
pull
content
from
this
repose
and
push
translated
content
to
this
repos
and
that's
how
it
works
and
fro
as
much
as
they
know,
everything
handed
by
crowd
inside
and
we
have
like.
We
can
subscribe
to
some
web
hooks
and
but
I
matures
at
sea
useful
because
there's
two
problems
so
first
problem
is
how
we
can
connect
this
model
front
and
take
the
list
of
strings
through
to
translate
it.
A
We
are
like
Hello
to
translate
for
machines
and
another
question
how
we
can
say
to
crowd
in
to
translate
only
things
these
strings
only
these
drinks.
So
this
is
like
two
questions:
requests
to
model
front
and
filter
for
crowd
in
so
probably
I'm
going
to
talk
with
scrolling
team,
because
with
some
support
yet
because
they're,
ready
and
I
will
ask,
does
it
possible
then,
because
we
need
like
to
figure
it
out?
Is
it
possible?
Probably
we
can
use
some
custom.
A
B
A
B
B
B
B
So
we
do
have
a
way
that
we
don't
expose
this
in
our
public
API.
So
we
can
think
about
it.
If
there's
a
demand
for
it,
of
course,
we'll
add
it,
it's
not
an
issue,
but
I
think
it
will
make
more
sense
because
you
know
essentially
would
be
resending
resending
resending
every
time
right.
The
whole
thing
with
slight
updates,
so
I
think
it
would
make
the
most
sense
to
only
send
the
changes
in
some
way.
Probably.
A
B
B
It's
getting
machine,
translated
right
away
and
then
any
any
any
volunteer
or
any
you
know
in
their
case,
not
volunteer.
Any
translator
who
sees
it
will
see
the
segment.
It's
like
getting
a
job
kind
of
right
and
then
sees
it,
and,
and
it
comes
with
the
machine,
translation
all
ready
to
start
post
editing
in
order
to
be
faster
and,
of
course
not
just
one
translation,
but
typically
right,
because
the
commit
is
a
file
or
something
it
will
be.
A
logical
chunk
has
some
context.
B
A
So
let
me
just
to
let
me
just
contract
to
crowd
in
team
and
discuss
a
possible
way.
Probably
there
have
some
API
events
for
API
web
hooks
for
session
memory,
update
and
probably
it
sends
strings
in
some
format
and
we
can
send
it
to
model
front
so,
okay,
first
guy
I
am
going
to
contact
with
crowd
in
team
and
then
I'm
going
to
back
to
you.
But
from
your
side
it
will
be
great
if
you
can
send
some
dogs
if
it's.
If
you
have
something
for.
A
Just
yeah
I
don't
take
a
look
and
because
yeah,
the
main
point
is
no
J's
team
would
like
to
use
some
automated
stuff
instead
of
manual
stuff
and
yeah,
so
I
need
to
investigate
because,
like
we
have,
we
have
okay.
Thank
you.
So
we
need
to
discuss
with
everyone
and
think
about
implementation
to
to
make
some
steps
now.
Iii
can't
say
something
definitely
yeah.
Well
yeah.
Thank
you
for
your
great
demo.
It
was
great,
and
now
it's
like
clear
most
of
the
points
so
yeah
thank.
A
Okay,
so
yeah
we
are
going,
we
have
just
20
minutes,
so
let's
go
quickly.
Unfortunately,
we
haven't
a
lot
of
members
on
the
call,
but
I
should
go
just
to
explain
each
of
the
points
so
yeah.
We
have
a
lot
of
stuff
like
open
it
trigger
at
this
week
and
we
are
not
very
active,
like
we
haven't
a
lot
of
problems
with
that.
A
The
key
point
is
related
to
crowding
formatting
stuff
is
annoying
and
we
are
I
am
connected
with
scrolling
team
and
they
are
promised
that
they
are
going
to
fix
this
soon
and
and
then,
if
you,
this
issue
are
going
to
be
resolved,
but
they
have
a
few
issues
related
to
formatting
and
it's
like
hard
and
I
hope
if
you'll
be
fix
it
soon
and
we
can
close,
but
for
now
it's
still
open.
So
I
just
want
to
warn
everyone.
All
translators
can
continue
their
translations
and
it
will
be.
A
The
process
is
the
same,
but
it's
for
developers.
Side
is
for
me
and
everyone
who
want
to
join
you
just
week
out.
Probably
someone
looking
for
updates
or
sugar.
Another
side
is
protect
folder,
with
translated
content
from
metal
updates
is
an
issue
that
we
discussed
before
is
about
prevent
manual
updates
for
our
content,
because
we
are
want
to
fully
movie
great
to
crowd
in
protest
instead
of
github
PR
flow,
and
we
need
to
protect
in
some
way
to
prevent
making
manual
changes.
A
And
here,
if
you
discuss
the
idea
first
one,
he
is
related
to
write
github
action
workflow
to
write
some
comments
on
there,
PR
that
will
change
it.
Content
and
I
started
working
with
that.
It's
like
you,
can
see
it
here.
This
is
PR
contain
one
file.
It's
called
protect
content
from
my
no
changes
is
checking
for
opens
pull
request
it
does
it
open
for
does
it
contain
changes
for
content
folder
and
if,
if
yes,
it
should
create
comment,
but
we
have
some
points
against
its
related
to
this.
A
This
event
probably
never
be
triggered
from
for
PR
sets
created
from
forks,
and
this
is
like
a
big
issue,
and
today,
I
have
some
time
to
research
and
I
I
discovered
that
this.
That
is
not
probably
a
case
because
we
are
using
public
repos
and
all
issues
that
I
found
is
related
to
private
is
private
people's,
so
I
suppose
we
can
try.
But
if
not,
we
need
to
find
another
way
for
how
we
can
like
protect
from
PRS,
and
one
more
thing
is
proposed
by
Zeke.
He
is
to
create.
A
Husky
github
webhook
sorry
github
cook
by
husky
to
write
some
message:
if
translator
trying
to
commit
something
in
in
the
same
folder
like
content
or
other
folder,
like
hey
man,
you
are
trying
to
translate
this
content
manually.
Please
use
crowding,
but
we
are
not
forcing
them
to
unknot
commit
some
something,
and
we
are
just
write
a
message.
It
should
be
polite
but
yeah.
A
This
is
an
idea
and
I'm
going
to
implement
this
on
the
next
thing,
because
this
week
was
really
crazy
and
yet
so
it's
like
quick
update
about
this
another
stuff
is
related
to
supported
versions,
versions,
strategy
and
yeah.
It's
really
a
low-end
stuff
that
I
discovered
on
the
previous
week
that
we
are
removing
all
content
related
to
duplicated
versions
or
like
end-of-life
versions.
We
are
removing
it
like
hard
remove
from
our
ripples,
and
this
never
exist
here
but,
like
don't
worry,
it's
still
contained
and
crowd
inside.
A
So
we
are
not
lost
it
forever
and
we
can
revert
it
and
everything
going
back.
But
I
just
figured
out
that
we
removing
this
from
repo
and
we
still
not
understand
the
build
process
and
we
still
not
agreed
some
ways.
How
we
can
deliver
our
translations
to
the
docs
website
or
just
to
usual
website
and
removing
their
content
is
like
bad
bad
way.
A
Them
on
actual
versions
and
textual
translations,
but
I
explained
that
crowding
has
priorities,
functionality
and
I
created
I,
updated,
crowding
guide
with
explanations
to
that.
So
now
it's
like
explained-
and
we
just
need
to
complete
with
versions
stuff,
to
revert
everything
so
I
updated.
Oh
yeah,
so
I
created
a
pull
request
that,
where
I
fetched
all
versions
from
force,
I
believe,
sir
from
version
number
four,
and
until
current
and
yeah
here,
I
also
played
two
tests
and
yeah.
So
it's
still
waiting
for
review
I,
don't
remember
what
the
requested
changes.
Oh.
A
Yeah
I
already
fixed,
but
Nick
still
not
make
another
review
yeah.
So
it's
open
and
I
am
waiting
for
resolving
this
PR
and
then
this
issue
will
be
resolved
and
we
are
stuck
containing
all
the
versions
on
our
side,
also
after
this
versions
will
be
merged
protein
scene.
This
changes
and
protein
slated
with
our
translation
memory
and
we
have
like
almost
translated
the
breaker
duplicated
versions.
A
A
Yep,
so
this,
like
about
this
point
yeah
another
stuff,
is
related
to
hiking
streams,
from
translation,
so
before
I
thought
that
we
that
crowd
Dean,
allow
us
to
provide
some
glossary,
and
this
Missouri
is
like
protect
some
strings
or
some
words
from
translations
like
so
no
translate,
no
GIS
or
not
gonna
translate
Windows
is
we
have
them,
but
it's
not
and
rhodium
don't
provide
this
functionality
with
the
glossary
and
this
functionality
possible.
Just
with
heights
drinks,
functionality,
yeah
I
conducted
200,
didn't
support
team
and
their
proposed
me.
A
This
solution
and
here's
like
again
question
about
automatisation
and
how
we
are
developers
and
we
should
automate
the
cultivator
everything
but
I'm
sure
this
is
possible
in
this
case.
So
here
is
a
quick
video.
How
we
can
do
this
on
crowd
in
I
can
show
you
just
quickly
here,
so
we
have
strings
for
each
project.
A
For
example,
we
don't
want
to
translate
node,
yes
and
we
just
search
all
it
shouldn't
be
searched
okay,
so
we
can
just
select
this
string
and
click
hide,
and
this
this
drink
will
never
be
shown
for
translators.
If
you
we
are
going
to
I,
don't
know
where
to
okay.
I
know,
one
case:
how,
where
I
I
prevented
from
translation?
If
you
link
so
I
just
going
to
show.
A
Yeah,
so
here's
a
list
of
strings
that
like
have
this
badge
hidden
and
it's
possible
to
click,
Save
and
I,
suppose
yeah
it's
possible,
but
it's
likes
hidden
and
it
should
be
like
drinks
for
Crusaders
to
not
translate
it
so
yeah,
it's
how
it
should
it.
It
can
be
handled
metal
as
I
show
you
now
and
it
also
can
be
handled
by
API.
A
But
from
my
point
of
view,
it's
really
hard
to
implement
a
bi
solution,
because
we
need
to
think
about
a
lot
of
stuff
related
to
how
we
decide
which
string
to
translate,
which
string
to
not
translate
that
think
about
patterns.
Think
about
some
projects,
I,
don't
know
and
from
I
said
it's
overhead
yeah.
Unfortunately
English
we
should
do
some
manual
stuff
instead
of
try
to
automate
it
yeah.
B
I
mentioned
in
my
notes,
was
that
the
benefit
of
this
kind
of
do
not
translate
least,
of
course,
how
it
gets
applied,
is
another
story,
but
the
benefit
of
it,
and
why
I
personally
think
it's
justified
as
an
investment
is
because
it
it
works
for
all
your
languages
yeah,
it
works
for
all
languages,
yeah.
Well,
it's
quite
a
good
payoff
and
unfortunately,
the
way
that
neural
machine
translation
is
going
right.
Now
this
kind
of
stuff
is
getting
translated.
B
But
it
has
this
problem,
so
I
would
I
would
just
start
investing
in
these
huge
lists
and
think
about
maybe
ways
of
automating
it
like
say:
okay,
I,
take
the
entire
node,
all
the
node,
we
go
vote
on
Gustav
Gustav
and
you
guzzled
are
something
like
the
buffer
read
file.
Sync,
all
of
these
kind
of
yeah
and
and
just
create
a
huge
list
of
those
and
say
okay.
This
is
all
there
better,
be
a
really
good
reason
to
translate
any
of
these
yeah.
A
A
Both
of
these
options
is
working
for
us
because
we
have
no
GS
crowd
in
bot,
which
is
not
real
user,
but
this
is
users
that
we
are
using
for
integrate
with
crowded
and
I
suppose
we
can
use
it
for
the
same.
Read
this
reasons.
But
again,
this
question
is
still
open.
Here's
as
you
can
see
it
here,
just
my
opinions,
my
comments
so
I'm
waiting
for
other
opinions-
oh
and
yeah,
so
I
just
proposed
two
ways
and
if
somebody
else
have
an
idea,
welcome.
A
Yep,
one
more
thing
is
open
it
by
unique
related
to
yeah.
It's
like
how
we,
how
our
translation
is
going
to
be
life
and
yeah.
Here's
some
discussion
in
another
repo
about
this
so
like
this
is
still
open
question
because
it
never
be
configured
to
deliver
translations
to
website
and
we
need
to
start
thinking
about
it.
I
hope
weak
start
started,
but
there's
no
any
marks
I
just
asked
it
can
I
help
with
something
and
then
waiting.
A
So
because
we
are,
we
have
some
conflicts
before
I
implement
and
he
implements
the
same
stuff
and
like
we
should
close
one
of
the
PRS
and
not
worry
very
comfortable
I.
Think
so,
like
it's
open,
question
and
I
think
we
need
to
start
thinking
about
building,
but
again
it's
not
related.
It
should
shouldn't
affect
to
our
translators.
Translators
still
can
continue
doing
their
great
stuff,
great
contribution
and
yeah.
A
It's
just
ready
to
build
in
can
just
continue
doing,
killer
stuff,
help
us
and
yeah,
and
one
more
thing
is
in
one
of
the
times
that
I
asked
it
crowding
support
for
help.
They
are
also
asking
me
about:
can
they
are
at
no
GS
logo
to
crowd
in,
as,
like
our
clients
and
I
said?
Okay,
me
I
think
there's
no
any
problems,
but
as
I
can
see,
it's
like
some
problems
here
exist
and
it's
it's
not
a
question
to
us
is
question
to
some
community.
You
committee,
a
team
and
click.
A
Let's
wait
for
their
response,
but
I
just
described
because
I
should
so
in
the
end
the
last
one
I
already
described
is
we
have
some
formatting
issues
which
crowd
in
and
throw
them
work
in
there
yeah
so
guys
any
question
I
I
did
big
monologue
yeah.
If
any
questions
here
we're
almost
done
with
our
time.
B
A
Okay,
so
no
any
questions
and
I
think
we
can,
and
with
this
making
yeah,
it
was
really
great.
Thank
you
for
demo
Adam
again
yeah,
it's
really
great,
and
now
it's
more
clear
for
for
me,
especially
yeah
and
I,
am
going
as
I
said,
I'm
going
to
investigate
with
this
process
and
the
standard
deeper
and
try
to
find
the
solution.
Who
is
crowd
in
and
try
to
integrate
with
you
yeah.