1
00:00:00,167 --> 00:00:10,626
...

2
00:00:12,834 --> 00:00:16,083
- Limsi is
a pluridisciplinary research lab

3
00:00:16,417 --> 00:00:17,459
of CNRS.

4
00:00:17,792 --> 00:00:21,125
Among its research fields,
there is everything about

5
00:00:21,459 --> 00:00:23,626
the sciences and technologies

6
00:00:23,959 --> 00:00:28,500
of languages, whether they are spoken,
written or signed.

7
00:00:29,334 --> 00:00:32,334
Limsi is more specifically
involved in subprojects

8
00:00:32,667 --> 00:00:35,667
that are about the production
of language resources,

9
00:00:35,999 --> 00:00:38,250
or corpus, and the development
of modules

10
00:00:38,584 --> 00:00:40,209
of automatic treatment.

11
00:00:40,542 --> 00:00:44,125
The goal is to automatize
the production of subtitles

12
00:00:44,459 --> 00:00:47,375
adapted to deaf
and hard of hearing people

13
00:00:47,709 --> 00:00:52,584
and also to produce a translation
in LSF, French Sign Language,

14
00:00:52,918 --> 00:00:55,584
for deaf people who speak it
as their first language

15
00:00:55,918 --> 00:01:00,250
and for whom it will be more accessible
than subtitling.

16
00:01:00,626 --> 00:01:03,999
We are more specifically involved
in sub-projects 3 and 5.

17
00:01:04,292 --> 00:01:08,626
Subproject 3 is a module
that is going to take as input

18
00:01:09,167 --> 00:01:12,500
a television show in the form of video

19
00:01:12,834 --> 00:01:16,626
and is going to produce captions.

20
00:01:16,959 --> 00:01:20,834
Subproject 5 takes
this caption

21
00:01:21,125 --> 00:01:24,250
and produces an LSF translation

22
00:01:24,584 --> 00:01:27,167
in the form of an animated avatar.

23
00:01:27,500 --> 00:01:30,584
So far, we have done basic bricks.

24
00:01:31,542 --> 00:01:34,667
For subproject 3, it's essentially

25
00:01:34,999 --> 00:01:39,542
a brick that allows
the transcription of speech

26
00:01:39,876 --> 00:01:42,250
from a television program video,

27
00:01:42,584 --> 00:01:46,751
and also a brick that starts
from this transcript

28
00:01:47,042 --> 00:01:49,792
to make a simplified version

29
00:01:50,792 --> 00:01:52,626
in the form of subtitles.

30
00:01:52,959 --> 00:01:56,250
Subproject 5 is
a much more exploratory work,

31
00:01:56,584 --> 00:01:59,667
the automatic treatment
of sign languages

32
00:01:59,999 --> 00:02:02,042
being much more recent.

33
00:02:02,375 --> 00:02:05,500
So we basically participated
in making a sign language corpus

34
00:02:05,834 --> 00:02:08,959
that will be used

35
00:02:09,250 --> 00:02:11,042
to generate animations.

36
00:02:11,375 --> 00:02:15,999
And then also, we have developed
an intermediary representation

37
00:02:16,292 --> 00:02:22,375
that is going to be a pivot between
subtitling and avatar animation control.

38
00:02:22,709 --> 00:02:26,584
Now, we are going to complete
and put together

39
00:02:26,918 --> 00:02:29,292
these basic bricks to have modules

40
00:02:29,626 --> 00:02:33,250
that work from end to end, and then
we are going to evaluate these modules

41
00:02:33,584 --> 00:02:36,876
and integrate all of them
into a final demonstrator.

42
00:02:37,167 --> 00:02:42,167
SYSTRAN