ARC – Using AI and ML to make media more accessible

VIDEO TRANSCRIPT

1
00:00:00,000 –> 00:00:04,400
So, I was at a recent thought leaders
forum on the future of media,

2
00:00:04,440 –> 00:00:08,880
and one of the key messages was,
“It’s all about reach.”

3
00:00:08,920 –> 00:00:11,840
Now, at Red Bee Media
Access Services,

4
00:00:11,880 –> 00:00:17,200
we create accessibility for reaching
people with sensory impairments.

5
00:00:17,240 –> 00:00:22,040
That includes closed captions –
subtitles in the same language,

6
00:00:22,080 –> 00:00:24,320
typically for the deaf
and hard of hearing.

7
00:00:24,360 –> 00:00:27,280
But it’s not just about
sensory impairments,

8
00:00:27,320 –> 00:00:29,880
it’s about anyone
who can’t hear the media –

9
00:00:29,920 –> 00:00:32,840
that’s if you’re in a noisy
environment, like a bar or a gym,

10
00:00:32,880 –> 00:00:35,840
or if your home is a noisy place
and you want to be able to watch

11
00:00:35,880 –> 00:00:40,920
the news with children running about
and high levels of noise like that.

12
00:00:42,320 –> 00:00:45,760
One of the big challenges around
closed captioning at Red Bee…

13
00:00:45,800 –> 00:00:49,400
We provide high quality
closed captioning and subtitles

14
00:00:49,440 –> 00:00:51,520
for a global range of audiences,

15
00:00:51,560 –> 00:00:55,280
for companies like the BBC
and the ABC in Australia

16
00:00:55,320 –> 00:00:57,640
and Spain and France and elsewhere.

17
00:00:57,680 –> 00:01:01,040
But reach
isn’t just about individuals,

18
00:01:01,080 –> 00:01:03,360
it’s also about getting range,

19
00:01:03,400 –> 00:01:06,040
making sure as many people
have accessibility

20
00:01:06,080 –> 00:01:08,800
across the wide range of media
that’s out there.

21
00:01:08,840 –> 00:01:12,440
The number of platforms and channels
that we consume media through,

22
00:01:12,480 –> 00:01:15,040
as we all know,
is increasing all the time,

23
00:01:15,080 –> 00:01:18,840
so we need to think about how
we make all that media accessible.

24
00:01:18,880 –> 00:01:22,360
Now, it may surprise you to know that
an awful lot of closed captioning

25
00:01:22,400 –> 00:01:24,600
is still very people-driven.

26
00:01:24,640 –> 00:01:28,920
We have expert closed captioning
teams delivering those services

27
00:01:28,960 –> 00:01:31,800
for those broadcasters and audiences.

28
00:01:31,840 –> 00:01:37,080
But could we reduce the cost and
increase the range of accessibility,

29
00:01:37,120 –> 00:01:41,680
to start addressing that much wider
media, and could machine learning

30
00:01:41,720 –> 00:01:46,200
and artificial intelligence provide
the solution for reaching that range?

31
00:01:46,240 –> 00:01:48,680
Well, that’s a fascinating
question for me,

32
00:01:48,720 –> 00:01:52,240
being involved in accessibility
now for nearly 20 years.

33
00:01:52,280 –> 00:01:55,040
So we set ourselves a challenge.

34
00:01:55,080 –> 00:01:57,920
Could we use machine learning
technology –

35
00:01:57,960 –> 00:02:01,120
artificial speech recognition
being one example of that –

36
00:02:01,160 –> 00:02:06,240
to drive a low-cost, high-quality
captioning solution that could be

37
00:02:06,280 –> 00:02:10,360
giving us accessibility in more
places than ever before?

38
00:02:10,400 –> 00:02:16,840
Increase the quality, decrease
the cost, increase the accessibility.

39
00:02:16,880 –> 00:02:21,040
Well, we thought we’d set
ourselves that challenge,

40
00:02:21,080 –> 00:02:23,600
so how did we go about it?

41
00:02:23,640 –> 00:02:27,120
We put together
a small, focused team,

42
00:02:27,160 –> 00:02:29,680
not just from
the technology space,

43
00:02:29,720 –> 00:02:32,080
but from across our range
of expertise,

44
00:02:32,120 –> 00:02:35,960
whether that’s the market-facing side
or our operational experts

45
00:02:36,000 –> 00:02:38,520
who are used to
delivering captions daily.

46
00:02:40,120 –> 00:02:42,240
And we got them and we sat down

47
00:02:42,280 –> 00:02:46,040
and we decided that we would just try
and forget everything we knew

48
00:02:46,080 –> 00:02:48,000
about subtitling and captioning,

49
00:02:48,040 –> 00:02:51,680
and forget everything we thought
we knew about machine learning,

50
00:02:51,720 –> 00:02:53,640
and start from scratch.

51
00:02:53,680 –> 00:02:57,080
We evaluated the market to see
which technologies were out there

52
00:02:57,120 –> 00:03:00,680
and we looked to see how we could
apply them successfully

53
00:03:00,720 –> 00:03:03,840
to deliver the sort of service
that people were expecting.

54
00:03:03,880 –> 00:03:06,240
The key thing was putting people
and audiences

55
00:03:06,280 –> 00:03:08,040
at the heart of what we did.

56
00:03:08,080 –> 00:03:11,840
That meant quick iterative loops
of feedback from audiences

57
00:03:11,880 –> 00:03:14,880
and our customers to make sure
we were doing the right thing

58
00:03:14,920 –> 00:03:18,160
with the technology in a way
that they would find accessible.

59
00:03:20,000 –> 00:03:21,240
The result?

60
00:03:21,280 –> 00:03:24,760
Well, last year we launched ARC
for our first customer

61
00:03:24,800 –> 00:03:27,200
in the United States, a PBS,

62
00:03:27,240 –> 00:03:30,720
public service broadcast station,
WLVT,

63
00:03:30,760 –> 00:03:35,680
and we launched it successfully
with them.

64
00:03:35,720 –> 00:03:39,000
We’ve just launched
our second customer in Spain,

65
00:03:39,040 –> 00:03:40,560
and the feedback has been fantastic.

66
00:03:40,600 –> 00:03:43,240
It’s been great from the customer,
who finds it easy to use

67
00:03:43,280 –> 00:03:45,080
because it’s automated,
and it’s been great

68
00:03:45,120 –> 00:03:47,760
from the audiences as well,
who are getting accessibility,

69
00:03:47,800 –> 00:03:50,680
and it’s at a cost
that everyone can manage.

70
00:03:50,720 –> 00:03:54,400
So, just to summarise,
good, tight team,

71
00:03:54,440 –> 00:03:57,080
forget everything you think you know,

72
00:03:57,120 –> 00:04:00,000
put audiences and people
at the centre of what you’re doing

73
00:04:00,040 –> 00:04:01,320
with machine learning,

74
00:04:01,360 –> 00:04:04,040
and that’s how you get
a great solution like ARC.

75
00:04:05,520 –> 00:04:07,320
So, I mentioned at the beginning

76
00:04:07,360 –> 00:04:10,680
that closed captions are one of those
things that we may all have used

77
00:04:10,720 –> 00:04:13,600
and seen, but have we really
thought about how they’re produced?

78
00:04:13,640 –> 00:04:15,840
Now, I think the world
is full of things like this,

79
00:04:15,880 –> 00:04:17,800
and we should remember
that machine learning

80
00:04:17,840 –> 00:04:20,240
and artificial intelligence
is transforming those things

81
00:04:20,280 –> 00:04:22,200
as much as
the higher technology things

82
00:04:22,240 –> 00:04:24,440
we may think about,
or the more obvious areas.

83
00:04:24,480 –> 00:04:27,200
But the important message
is that we need to put people

84
00:04:27,240 –> 00:04:29,720
at the centre of working
with these technologies,

85
00:04:29,760 –> 00:04:32,320
whether it’s for people
or the people who are working with

86
00:04:32,360 –> 00:04:35,280
the machine learning technology
themselves.

87
00:04:35,320 –> 00:04:39,640
It’s people with machine learning,
not people against machine learning.

88
00:04:39,680 –> 00:04:42,360
So the next time you see
closed captions on the screen,

89
00:04:42,400 –> 00:04:45,000
those words at the bottom
of the screen to help the deaf

90
00:04:45,040 –> 00:04:48,360
and hard of hearing, have a think
about how they’re produced,

91
00:04:48,400 –> 00:04:51,560
and remember, ultimately,
it’s all about reach.

Share Twitter Linkedin

AI and ML are much more than trending buzz words. When innovating in Red Bee, we use the latest technology together with our people’s expertise to make exciting products, like ARC (Automatic Remote Captioning), that helps the media industry to extend the reach of what they do.

So we set ourselves a challenge: “Could we use machine learning technology, artificial speech recognition being one example of that, to drive a low-cost, high-quality captioning solution that could be giving us accessibility in more places than ever before?”

The objective was: Increase the quality, decrease the cost, increase the accessibility. The key thing was putting people and audiences at the heart of what we did and last year we launched successfully ARC for our first customer in the US.

So the next time you see closed captions on the screen, those words at the bottom of the screen to help the deaf and hard of hearing, have a think about how they’re produced, and remember, ultimately, it’s all about reach.

Share Twitter Linkedin