Impacts of MT and social media on language services
1. Transla'on
in
21st
Century
Impacts
of
MT
and
social
media
on
language
services
Rahzeb
Choudhury
9
February
2011
2. T21:
key
themes
Content
explosion
–
self
evident
Content
disrup'on
–
Some
created
by
companies
…most
created
by
users….oDen
social
media,
add
mul'media
(audio,
video,
anima'on),
embracing
“good
enough”
(prac'cal
is
more
important
than
preKy
and
perfect)
Content
differen'a'on
–
Assessing
aKributes
of
u'lity,
'meliness,
sensi'vity
Technology
–
open/interoperable,
collabora've,
machine
transla'on
Transla'on
as
a
u'lity…
3. Standards
and
interoperability
LISA
-‐TAUS
survey,
Jan-‐Feb
2011
93
companies
to
date
Define
interoperability
Op0ons
%
responded
Standard
format
for
exchange
of
76
terminology
Standard
format
for
exchange
of
transla'on
77
memory
Standard
for
interac'on
of
content
75
management
systems
with
transla'on
management
systems
and
MT
Standard
business
process
21
4. Standards
and
interoperability
LISA
-‐TAUS
survey,
Jan-‐Feb
2011
93
companies
to
date
Conclusion:
Material
issue!
Lack
of
interoperability:
cost
to
your
business
%
responded
<5%
10
Between
5-‐10%
12
Between
10-‐20%
14
Between
20-‐30%
12
More
than
30%
14
Don’t
know
40
5. Standards
and
interoperability
LISA
-‐TAUS
survey,
January-‐February
2011
Ac'on:
Surveillance
93
companies
to
date
of
compliance?
Biggest
barrier
Average
ra0ng
Lack
of
compliance
to
interchange
formats
3.08
Few
suppliers
dominate
the
market
2.97
No
capable
umbrella/organizing
body
2.96
Apathy
2.77
Industry
is
immature
2.75
No
budget
to
deal
with
problem
2.65
6. MT
and
the
Dutch
market
ATA-‐TAUS
survey,
December
2010
31
companies
18
companies:
Most:
revenue
> revenue
<€500k
pa
€1m
pa
We
operate
MT
inhouse
and
provide
We
do
not
PE
provide
PE
13
companies
–
19%
74%
not
suitable
for
the
content
we
We
regularly
translate
provide
PE
7%
10
–
We
are
looking
into
providing
PE
7. MT
and
the
Dutch
market
ATA-‐TAUS
survey,
December
2010
Sample:
31
companies
MT
Users
(#6)
1
uses
Moses,
1
licenses
soDware
and
is
trialing
Moses,
4
license
MT
vendor
soDware
Pros
-‐
Speed,
efficiency,
compe''ve
advantage
Cons
-‐
Customiza'on
'me,
quality
of
output
not
always
up
to
scratch
We
go
deeper
with
presenta0ons
from:
Ge7ng
started
with
MT:
Nathalie
De
Su>er,
Cross
Language
Trialing
Moses:
Raymund
Prins,
Global
Textware
Managing
source
text
:
Bert
Bourgonje,
L&L
MT
and
client
requirements:
Wayne
Bourland,
Dell,
and
John
Dixon,
ALS
8. European
MT
providers
European
proprietary
MT
vendors:
MorphoLogic,
Linguatec,
Lucy
SoDware,
SynTHEMA,
Systran,
Ta
with
you
…..most
took
many
years
and
millions
to
come
to
market
European
SMEs
with
open
source
based
solu'ons:
Applied
Language
Solu'ons,
Cross
Language,
Eleka,
LanguageLens,
Pangeanic,
Promsit
Language
Engineering,
Simple
ShiD,
Translated.net…..came
to
market
more
recently,
much
faster
and
cheaper.
A
couple
are
growing
rapidly…
Something
even
more
radical
–
presenta0on
by
TranslaNon
by
the
hour:
David
Sowerby,
Straker
SoPware…example
of
a
new
innovaNve
market
entrant…
9. How
long
ago
did
you
first
download
the
Moses
Decoder?
Less
than
1
month
ago
1-‐3
months
ago
3-‐6
months
ago
6-‐12
months
ago
7%
9%
11%
55%
18%
Source:
EuroMatrixPlus-‐TAUS,
Jan-‐Feb
2011
Sample:
45
organiza'ons
to
date
10. We
are
in
the
following
phase:
0%
Pilo'ng
4%
Implementa'on
Produc'on
33%
47%
Not
using
Moses
Not
applicable
(please
explain
why
briefly)
16%
Source:
EuroMatrixPlus-‐TAUS,
Jan-‐Feb
2011
Sample:
45
organiza'ons
to
date
11. Main
purpose
of
survey
to
focus
developments
on
main
areas
for
improvement.
Respondents
advise:
1.
Needs
to
be
easier
to
integrate
into
workflow
(A
solu'on
-‐
Moses4Localiza'on
–
available
March
2011)
2.
Faster
training
and
transla'on
speed
(Esp
for
online
service
-‐
Number
of
projects
underway
-‐
expect
significant
improvements
within
18
months
or
earlier)
3.
Wider
range
of
languages
where
sa'sfactory
results
(This
is
the
main
focus
of
academic
research)
4.
Needs
to
be
easier
to
install
and
use
(Expect
improvements
within
6
months)
Source:
EuroMatrixPlus
and
TAUS,
Jan-‐Feb
2011
Sample:
45
organiza'ons
to
date
12. MT
adop'on
–
on
course
for
B...or
C
(thousands
of
MT
systems)
Scenario C
MT
providers
Scenario B –
two dozen
Time
providers
Scenario A –
very few large
players dominate
MT
customers
14. Clients
would
ideally
like
a
passionate
language
community
Why?
• A
boost
to
transla'on
resources
• Can
improve
brand
percep'on
• Enhance
user
experiences
• Engage
people
to
get
feedback
15. Approaches
to
community
transla'on
•
Social
network
members
becoming
the
community
of
translators
•
Virally
recruited
and
self
organizing
crowds
•
Communi'es
of
product
users
who
benefit
directly
from
the
transla'on.
ODen
recognized
as
authors.
Most
oDen
used
for
second
'er
languages,
where
there
is
a
shortage
of
translated
material.
The
process
is
managed
by
and
postedi'ng
done
by
vendors
•
Communi'es
who
postedit
MT
output
for
non
commercial
content
• Open
source
product
communi'es
with
ideological
mo'va'ons
•
Internal
employee
communi'es,
such
as
for
in-‐house
content
that
is
not
postedited
•
In
B2B
sectors,
resellers/partners
who
act
as
the
community
in
smaller
markets
to
benefit
from
increases
in
sales
revenue/barter
arrangements
and
in
some
cases
share
support
revenue
16. Two
approaches
to
quality
assurance
• The
first
leaves
it
to
the
crowd.
Here
there
may
be
awards,
group
vo'ng,
peer
review,
rankings
and
some'mes
a
general
acceptance
that
usefulness
rather
than
publishable
quality
is
the
required
benchmark
• The
second
approach
uses
a
valida'on
stage
involving
professional
linguists
who
postedit
17. Issues
• Difficulty
in
pressuring
a
crowd,
say
to
meet
a
deadline
• Difficulty
in
controlling
file
standards
and
benefits
to
the
TM
database
• Firewall
issues
which
are
solve
d
by
transla'on
taking
place
in
discussion
forums
18. The
toolkit
“Unlimited
user
licenses,
terminology
management,
clean
and
intui've
transla'on
interface,
transla'on
sugges'ons,
automa'c
TM
updater,
MT
sugges'ons,
aliases
and
IDs,
admin
rights
for
language
managers,
in-‐
context
transla'on
workbench,
simple
debugger
and
resource
editor.
“
Stephan
Cocron
of
VeriSign
Inc.