Captioning webinarpart1

							1
00:00:06,900 --> 00:00:11,666
>> Soji:
Hi. Welcome to Captioning in the University Environment Seminar.     First,
I would like to

2
00:00:11,666 --> 00:00:14,632
introduce our Accessing Higher Ground conference

3
00:00:14,633 --> 00:00:19,599
>> Soji:
coordinator. Mr. Howard Kramer for University of Colorado at Boulder. He
will be our

4
00:00:19,600 --> 00:00:25,933
moderator of today's webinar. Howard, will you press on the become
presenter button please and start our presentation?

5
00:00:25,933 --> 00:00:29,199
>> Soji:
  Thank you very much.

6
00:00:29,200 --> 00:00:37,133
>> HOWARD KRAMER:
Thank you, Soji. And welcome to everyone who is connected in.   My name
is Howard Kramer.

7
00:00:37,133 --> 00:00:43,399
>> HOWARD KRAMER:
Again I am at the University of Colorado and this is our -- welcome to
webinar "Captioning in the

8
00:00:43,400 --> 00:00:52,566
University Environment." This is the first webinar we have conducted. So
we are excited about it and also a little nervous. So we will see how it

9
00:00:52,566 --> 00:00:59,999
goes. Some of you may be familiar with the university and with me from
the Accessing Higher Ground Conference.

10
00:01:00,000 --> 00:01:07,800
And that occurs every November.   And we hope this to be the first in a
series of teleconferences or

11
00:01:07,800 --> 00:01:14,766
>> HOWARD KRAMER:
webinars that we put on in between conferences.   So, you know, when we
put this on this seems like

12
00:01:14,766 --> 00:01:17,266
 it must be a very hot topic right now.

13
00:01:17,266 --> 00:01:24,099
>> HOWARD KRAMER:
I was hoping for ten participants I thought would be good.   We have
gotten 35 different institutions

14
00:01:24,100 --> 00:01:30,466
or at least 35 different sites are logged in to this seminar today.

15
00:01:30,466 --> 00:01:37,432
>> HOWARD KRAMER:
So my goal today is actually to talk as little as possible and just
provide you with the information

16
00:01:37,433 --> 00:01:41,566
you need and to let the presenters share their expertise with you.

17
00:01:41,566 --> 00:01:51,466
>> HOWARD KRAMER:
I just wanted to go over a few logistics and a few housekeeping issues
before we go on. The first is

18
00:01:51,466 --> 00:02:07,066
the -- just some basic help if problems occur. And let me bring up my
next power point. So if you have any problem with the V-cube

19
00:02:07,100 --> 00:02:14,833
connection which is the -- which is the service provider providing the
webinar interface,

20
00:02:14,833 --> 00:02:30,299
you can call their number that's up on the screen, 310-329-5959,
extension 8008. If you have captioning issues you can use chat to
indicate

21
00:02:30,300 --> 00:02:41,133
this. You can -- or you can e-mail Jill Perry, my colleague, at
jill.perry@colorado.edu. And that email is up on the screen. If you've
got questions

22
00:02:41,133 --> 00:02:45,366
you can put them in to the chat or you can e-mail

23
00:02:45,366 --> 00:02:56,999
me those questions. Obviously, the chat questions will probably be seen
by the presenter immediately. A little note that the chat is not

24
00:02:57,000 --> 00:03:06,066
public. So only the presenters see what's typed in.   And that's just to
allow some privacy

25
00:03:06,066 --> 00:03:14,066
between questions and the presenters. There is also -- I wanted to
direct people to the registrant resource page. This powerpoint and a lot
of other

26
00:03:14,066 --> 00:03:16,166
 resources are located on that page.

27
00:03:16,166 --> 00:03:27,232
And I e-mailed this URL last night and I just e-mailed it again to all
the people who are listed as the contacts for this conference. So you
can

28
00:03:27,233 --> 00:03:39,299
access that page and access these numbers and e-mails after the
presentation starts.

29
00:03:39,333 --> 00:03:51,599
So I wanted to go over the agenda for today. As I mentioned in my e-mail
last night it has changed slightly. Just a minor adjustment. So
starting at

30
00:03:51,600 --> 00:03:57,966
10:30 John Foliot from Stanford University will do -- will talk about
what the

31
00:03:57,966 --> 00:04:05,799
projects they are working on at Stanford .. basically talking about their
model for producing a sort of a turnkey environment for captioned
32
00:04:05,833 --> 00:04:16,566
media. That will be followed by 15 minutes -- let me backtrack -- John's
talk will be about an hour and will be followed by 15 minutes of
question

33
00:04:16,566 --> 00:04:23,999
and answer. We'll then we will take a break at 11:45 for 20 minutes and
then resume at 12:05.

34
00:04:24,000 --> 00:04:34,533
At that point Dean Brusnighan will do a short update of what's been
happening at Purdue University regarding captioning. Hopefully all

35
00:04:34,533 --> 00:04:40,499
of you have seen the prerecorded content. You'll get the most out of his
talk obviously if you've seen that.

36
00:04:40,500 --> 00:04:50,533
You can watch it after this webinar. It will be up and posted along with
everything that goes on today will be posted and available

37
00:04:50,533 --> 00:04:55,033
for at least a year for people who want to look at it afterwards.

38
00:04:55,033 --> 00:05:03,033
At 12:35, this is the new element of the agenda, I had actually been
talking to Angella Anderson at the University of Illinois

39
00:05:03,033 --> 00:05:12,766
about our conference in November and about what she was planning on
presenting and realized it fit really well in to today's topic. So I
asked her

40
00:05:12,766 --> 00:05:21,732
to just spend 15 minutes talking about what's going on at the University
of Illinois. We will use the last 30 minutes to talk to -- to provide a

41
00:05:21,733 --> 00:05:31,866
it will be an opportunity to ask questions to John, Dean or Angella.   And
we will just leave five minutes at the end for concluding remarks.

42
00:05:31,866 --> 00:05:52,066
And hopefully we will be able to fit everything in. Okay. I am going to
just go back once. So for those of you who are using captioning I just

43
00:05:52,066 --> 00:05:55,332
 wanted to give a tip on how to set that up on the

44
00:05:55,333 --> 00:06:08,133
screen. I found the best way is to since the captioning is being
provided in a separate URL, so a separate browser window so to speak, I

45
00:06:08,133 --> 00:06:17,333
found the best way to set this up is to put it -- minimize it as
indicated up on the powerpoint I have on the screen and putting it to the
right or left

46
00:06:17,333 --> 00:06:22,233
of the V-cube presentation. If you try to put it on the bottom it kind
of

47
00:06:22,233 --> 00:06:33,433
squeezes the V-cube interface too much and it is hard to read the
powerpoints. If you -- again if you have any problems with any of this
you can enter

48
00:06:33,433 --> 00:06:43,466
- type something in to chat or you can send me an e-mail. And again and
if you are having specific captioning problems you can also send an e-
mail

49
00:06:43,500 --> 00:06:59,933
to Jill Perry.   So I think at -- I will make one other suggestion and
that's I think you will find that the volumes vary. We did sound tests
yesterday and

50
00:06:59,933 --> 00:07:06,633
everything seems to be working very well. But I did notice that volumes
varied greatly. Some of our speakers are soft spoken and some are

51
00:07:06,633 --> 00:07:17,166
louder. So be prepared to adjust your volumes especially if you are
playing these through speakers. And with that I would like to hand over

52
00:07:17,166 --> 00:07:25,832
the session to our first speaker today, John Foliot from Stanford
University. John.

53
00:07:27,833 --> 00:07:33,199
>> JOHN FOLIOT:
Well, thank you, Howard. And good morning, everyone.   Yeah, I am John
Foliot. I work at the

54
00:07:33,200 --> 00:07:40,300
Stanford University. I run the Stanford online accessibility program and
the presentation I am going to do

55
00:07:40,300 --> 00:07:46,566
>> JOHN FOLIOT:
today is looking at the work flow model for captioning that I worked on
with my friend and

56
00:07:46,566 --> 00:07:59,532
associate Sean Keegan who works out of our office of accessible
education. So whoops, sorry. So this presentation is also available on
our

57
00:07:59,533 --> 00:08:08,833
>> JOHN FOLIOT:
website at Stanford. It is captioning.stanford.edu/ presentations.   It is
on line now.

58
00:08:08,833 --> 00:08:14,399
>> JOHN FOLIOT:
Most of it. With the blue/black background.There is a couple of slides
that I have added to the

59
00:08:14,400 --> 00:08:20,733
presentation for today that are not online but you can also contact me
directly if you need those slides as well.

60
00:08:20,733 --> 00:08:33,833
So, about three years ago we started to look at why captioning wasn't
happening on campus. What the barriers for getting captioning produced

61
00:08:33,833 --> 00:08:38,466
 were and I talked to a number of web masters

62
00:08:38,466 --> 00:08:48,799
and content producers on campus and we started asking why aren't you
getting this done and very quickly sort of came to a number of sort of

63
00:08:48,800 --> 00:08:51,366
 preconceptions and assumptions on their behalf.

64
00:08:51,366 --> 00:09:00,266
One of the things that I heard time and time again was that producing
captions was very costly to produce. That .. there was an expense that
was

65
00:09:00,266 --> 00:09:12,766
found to be onerous and that people just found it really expensive to
have this done. Many people also told me it was really geeky. It
required a fairly

66
00:09:12,766 --> 00:09:24,299
high level of technical expertise and knowledge to create captioning.
That the tools were not very intuitive, they were hard to work with. And
that

67
00:09:24,300 --> 00:09:32,066
there was a fairly steep learning curve to create captioned media. It
was also thought that there were perceived -- that it took a lot of time
to

68
00:09:32,066 --> 00:09:43,066
produce captioned videos. And for a lot of the people that were intrusted
with getting content up on the web quickly, they just, you know, one of

69
00:09:43,066 --> 00:09:52,799
the things I heard time and again is they didn't have the time to sit
down there and manually caption a video and a lot of people and for those

70
00:09:52,800 --> 00:10:02,166
of you that work in the accessibility space this is not an uncommon
refrain. A lot of people didn't really see the value of providing
captioning

71
00:10:02,166 --> 00:10:12,132
outside of accommodating deaf and hard-of-hearing users. That it was
strictly an accessibility thing. And while I don't have any deaf
students in

72
00:10:12,133 --> 00:10:20,466
my classes is the thing we always hear. So those were some of the basic
responses that I got when I started talking to people on campus.

73
00:10:20,466 --> 00:10:30,599
So after thinking about it and sort of analyzing some of these responses
Sean and I decided that what we really needed to do was to develop a

74
00:10:30,600 --> 00:10:40,733
workflow solution that would basically level out these bumps in the road
and make creating accessible new media something that was

75
00:10:40,733 --> 00:10:49,733
relatively easy to do and didn't require a lot of technical skill as well
as we realized that we had to do some work towards promoting the benefits

76
00:10:49,733 --> 00:10:57,866
of captioning and to explain why it was useful for users rather than --
above and beyond deaf and hard-of-hearing users.

77
00:10:57,866 --> 00:11:12,466
So for the first part what we did is we looked at a system that would be
as turnkey or as flow through as possible. So the basic premise is that
the

78
00:11:12,466 --> 00:11:22,932
content author uploads the video here. Soji, I don't know how to change
the color of the pen but they upload the video or audio file in to our
system and

79
00:11:22,933 --> 00:11:38,166
the first thing that the system does is it converts the video in to a
series of media files whether it is an FLV or a MP4 using HT-64 encoding,
MP3 for

80
00:11:38,166 --> 00:11:50,366
an audio file and WebM. But one of the things that it does is it also
takes the MP3 and it creates an audio file only even though it is a
video. To get

81
00:11:50,366 --> 00:12:01,266
captioning all we really need is the audio and the system outsources the
audio to a transcription vender who actually does the conversion of the

82
00:12:01,266 --> 00:12:12,199
speech-to-text. Once that text or once the audio is converted in to text
it is fed back in to the system where the system creates a series of

83
00:12:12,200 --> 00:12:23,633
caption files and again we are creating a number of different file
formats, DFXP or TTML with WC3 standard. We are also creating a SRT which
is

84
00:12:23,633 --> 00:12:35,033
kind of a loose coda or encoding standard, timestamp standard that really
-- it is not a standard per se but it is a specification that is

85
00:12:35,033 --> 00:12:45,799
widely used and supported by a number of third party media players. We're
also creating a binary file format called SEC or Senturis closed
caption..

86
00:12:45,800 --> 00:12:54,366
This is very similar to the line 21 caption files that are produced for
television. And it is very common in the broadcast media field.

87
00:12:54,366 --> 00:13:03,232
We discovered early on and less so now because there has been some new
tools that arrived on the market but when we started on this program this

88
00:13:03,233 --> 00:13:04,499
 was also a file format that

89
00:13:04,500 --> 00:13:13,833
was required to provide closed captions to the MP4 files that would be
used on your "i" Devices, the iPods, the iPhone,

90
00:13:13,833 --> 00:13:17,533
et cetera, and of course you also get back the text.

91
00:13:17,533 --> 00:13:31,966
So recently we also added a couple of different things to the system. So
the ability to bypass the video. So one of the things that we discovered
is

92
00:13:32,033 --> 00:13:44,133
that a lot of people were shooting in high definition video and our
system is not set up to actually support or create high definition video
for

93
00:13:44,133 --> 00:13:52,933
web delivery. So we created a system where you didn't actually get a
video file converted by the system but rather you could just upload the
video

94
00:13:52,933 --> 00:14:02,599
or we would do an audio conversion. So again remember we were creating
media files. The system will ingest a number of different video files

95
00:14:02,600 --> 00:14:11,566
in to the system and what we do is we output some sort of web ready
videos in these different formats but we couldn't support a high def
just

96
00:14:11,566 --> 00:14:23,299
because our system was unable to do that code conversion. So we created
a bypass for that. We also started providing support for the letter box
or

97
00:14:23,300 --> 00:14:26,166
the 169 video aspect ratio.

98
00:14:26,166 --> 00:14:35,532
Originally started off with the 43 which was sort of the common video,
sort of 800 x 600 kind of frame. But again more people are going

99
00:14:35,533 --> 00:14:43,299
for the larger sort of letter box format, what you see very often on your
iPhones and other devices like that.

100
00:14:43,300 --> 00:14:51,866

We also -- originally we were only outputting at one size of video.    So
we added a couple of

101
00:14:51,866 --> 00:14:56,999
different video resolution sizes.   Sort of acknowledging the fact that
websites and web pages

102
00:14:57,000 --> 00:15:04,600
>> JOHN FOLIOT:
are increasingly having more real estate just because end users desktop
screens are a lot

103
00:15:04,600 --> 00:15:14,700
larger and we started providing export option for WebM which is the VP8
Codec. A little bit more about that later in the presentation.

104
00:15:14,700 --> 00:15:28,033
But this is the latest Codec. It is an open source or patent free
encoding Codec. It is very directly linked to sort of HTML5 efforts with
video. And

105
00:15:28,033 --> 00:15:37,266
while HTML5 itself does not yet support native closed captions we were
getting requests from content producers on campus that

106
00:15:37,266 --> 00:15:46,366
they wanted to start working with WebM. There are a number of sort of
JavaScript solutions that can be used using WebM and caption files. So

107
00:15:46,366 --> 00:15:55,732
we are now producing content like that. So what I would like to do is
originally I was hoping to actually do a live demonstration but

108
00:15:55,733 --> 00:16:04,266
unfortunately we can't really do that. So I have done a series of screen
shots and I will walk you through the process of how

109
00:16:04,266 --> 00:16:17,532
content producers on campus can create captioned videos with a minimum of
effort. So we have a website that users go to. And before

110
00:16:17,533 --> 00:16:23,399
anybody on campus can start to use the system, they have to first
register with the system.

111
00:16:23,400 --> 00:16:35,866
So that we can track them and I mean as any user or system like this you
need to have a user profile. One of the things that's really important
in the user

112
00:16:35,866 --> 00:16:45,466
profile however is we are capturing billing information. And I will touch
on that a little bit more but one of the goals that Sean and I had when
we

113
00:16:45,466 --> 00:16:54,066
were setting up the system is we wanted this to be a pass through system.
We didn't want to have to be actively involved in maintaining and

114
00:16:54,066 --> 00:17:03,632
managing this system. And to that end right now we probably spend no more
than an hour or two a month doing maintenance on the system. So It is

115
00:17:03,633 --> 00:17:13,333
very much a pass through system. The captioning or the transcription
costs are borne directly by the people that are using the system.

116
00:17:13,333 --> 00:17:22,099
The transcription companies that we are working with are invoicing our
users directly. So we are completely out of sort of that billing and,
you know,

117
00:17:22,100 --> 00:17:24,000
 sort of financial management situation.

118
00:17:24,000 --> 00:17:34,500
So people need to register and Sean and I receive an e-mail when somebody
is looking to register for the system that we manually approve.

119
00:17:34,500 --> 00:17:44,666
At this point in time we have limited the number of people that can use
the system to people that are directly connected to the university with
one or two

120
00:17:44,666 --> 00:17:54,399
very minor exceptions. We are not looking to become an outsourcing
organization. We don't have the cycles or

121
00:17:54,400 --> 00:18:06,233
resources to support people off campus. The exception is my friend Victor
Tsaran who runs the accessibility lab at Yahoo. He himself is blind and

122
00:18:06,233 --> 00:18:17,433
we have a really strong relationship with that usability lab over at
Yahoo given that we are also located in the Bay area. So we open the
system

123
00:18:17,433 --> 00:18:26,666
up to him. He provided some early feedback to us as well in terms of how
the system works for nonsighted users. So he provided some sort of

124
00:18:26,666 --> 00:18:29,266
 real world experience and feedback to us.

125
00:18:29,333 --> 00:18:37,766
So you either register. If you have forgotten your password, you put in
the user, you click on here

126
00:18:37,766 --> 00:18:40,999
and you put in your e-mail address and you receive an e-mail that gives

127
00:18:41,000 --> 00:18:47,266
you a one time log-in so that you can log in and you can change your
password, et cetera, et cetera.

128
00:18:47,266 --> 00:18:57,932
So once you have logged in to the system, you are presented with a -- the
video upload page where you create a new project. So the first thing

129
00:18:57,933 --> 00:19:01,733
 you would do is you select your media file.

130
00:19:01,733 --> 00:19:10,199
There is a video conversion profile that you select and I will be going
through all of these in more detail later.

131
00:19:10,200 --> 00:19:20,266
The one thing that's critical is that you enter a title of the project
and the reason for that is you need a title so you can find stuff in your
library later on.

132
00:19:20,266 --> 00:19:31,366
You have the option of providing a brief description of your project so
that, you know, when you go back later on you will be able to sort of

133
00:19:31,366 --> 00:19:39,199
remember what this video was about. One of the things that's critical to
the system is the fact that we are

134
00:19:39,200 --> 00:19:50,600
allowing or we have set it up so that a number of third party
transcription companies are actually providing the speech-to-text
transcription.

135
00:19:50,600 --> 00:19:59,166
If anybody has played with captioned videos right now you will know that
getting your speech converted to text is the most time consuming and

136
00:19:59,166 --> 00:20:02,632
 difficult process, part of the entire process.

137
00:20:02,633 --> 00:20:11,899
And so we have contracted with a number of different transcription
companies in the Bay area and we went to them -- actually they are not
all in

138
00:20:11,933 --> 00:20:17,766
the Bay area. They are actually distributed throughout the country. They
are all North American based

139
00:20:17,800 --> 00:20:28,566
transcription companies. There are some organizations that outsource
their material to countries outside of North America.

140
00:20:28,566 --> 00:20:38,332
We were concerned about accuracy, given some of the technical terms that
we were dealing and just kind of the level of content, sort of
educational

141
00:20:38,333 --> 00:20:40,766
 content that we were looking at.

142
00:20:40,766 --> 00:20:49,999
So that was kind of one of our prerogatives.   You could use virtually any
caption or transcription company out there.

143
00:20:50,000 --> 00:20:56,366
We went to these companies and we asked them to give us a couple of
different price points based on turnaround time.
144
00:20:56,366 --> 00:21:06,532
So right now you can see I have two companies that are providing a 24-
hour or same day turnaround time to convert the speech-to-text.

145
00:21:06,533 --> 00:21:17,533
I have a number of companies that are giving it to me in the 24 or rather
48 hour time frame, two business day time frame. We were very specific

146
00:21:17,533 --> 00:21:20,433
 about defining it as business days.

147
00:21:20,433 --> 00:21:30,233
The last thing we wanted was somebody uploading a video at 5 o'clock on a
Friday afternoon and expecting to have it Sunday, you

148
00:21:30,233 --> 00:21:32,933
 know, in their inbox because frankly

149
00:21:32,933 --> 00:21:40,899
these companies, you know, they are commercial companies and they take
weekends as well. So we specified business days, four business days

150
00:21:40,900 --> 00:21:47,600
and a week. So depending on the turnaround time or in the case of this
organization, here Docsoft, the

151
00:21:47,600 --> 00:21:54,500
number of speakers that are included in the audio file, they have given
us a couple of different price points.

152
00:21:54,500 --> 00:22:06,466
And so we are agnostic. We really don't care which company our users
use.   What I can tell you is that right now Cogi at $1.25 a minute for a
48

153
00:22:06,466 --> 00:22:17,499
hour turnaround time seems to be getting the lion's share of the work
which tells us -- I am seeing some questions here.

154
00:22:17,500 --> 00:22:27,300
So somebody is asking does Stanford sell the system. We do not. Karen,
I will pull up my e-mail address at the end and I can provide you with
155
00:22:27,300 --> 00:22:32,300
some contacts. I'll tell you how the system was put together a little
later on. I can find you some contacts.

156
00:22:32,300 --> 00:22:49,100
Temple University, they are not getting any sound. I have to hand that
off to the V-cube people to try and figure out. So as I said the $1.25 a
minute

157
00:22:49,100 --> 00:22:51,966
turnaround time from Cogi seems to be

158
00:22:51,966 --> 00:23:01,232
the one that most people are using which again sort of confirms to me
that people are very price conscious about this. So finally one of the
things

159
00:23:01,233 --> 00:23:11,266
that we looked at when we were setting up the system was special terms,
special vocabularies. So like any large institution you are going to
have

160
00:23:11,266 --> 00:23:15,366
for example, buildings, or family names that you want to make sure

161
00:23:15,366 --> 00:23:27,366
that the spelling is correct. There are also legal terms, medical terms
and specific engineering terms. And so when the user is uploading

162
00:23:27,400 --> 00:23:35,933
a video to the system, if they are aware of some special terms, so, for
example, in the medical school there are a lot of medical terms that a
lot of

163
00:23:35,933 --> 00:23:41,999
transcriptionists would likely not be familiar with. So one of the things
that happens is you can add

164
00:23:42,000 --> 00:23:57,433
a special term that's associated to this video. And what happens is that
those terms are passed along to the transcription company via an e-mail.

165
00:23:57,433 --> 00:24:07,599
They are also added to a flat text file. And so the transcription
companies can log in, they can download that flat text file and
essentially it

166
00:24:07,600 --> 00:24:16,933
becomes a custom dictionary that they can incorporate in to their
transcription tools. So stepping through the process,

167
00:24:16,933 --> 00:24:25,033
selecting a media file. It is really simple. You click on the browse
button and you find the video file format that you are looking for.

168
00:24:25,033 --> 00:24:38,133
Right now we support MOV, Windows media files, WMA and MP4. We also
support audio only files of Wave and MP3. One of the problems that we

169
00:24:38,133 --> 00:24:47,099
have right now and it is a problem -- it is a limitation with the system
and we are working with some engineers to try and overcome it, but the

170
00:24:47,100 --> 00:24:55,200
maximum file size that you can upload in to the system right now is a
gigabyte. So again I mentioned earlier that we have some people who

171
00:24:55,200 --> 00:25:01,700
are shooting high def or high definition video and those files either due
to

172
00:25:01,700 --> 00:25:13,466
screen size or just the quality often exceed one gigabyte. So another
benefit of being able to upload the audio files I have two content
creators

173
00:25:13,466 --> 00:25:18,766

on campus who are doing a fair amount of high definition video right now.
So as part of their post

174
00:25:18,766 --> 00:25:30,999
production they are ripping a MP3 from their video and uploading the MP3
from their system. The next step is to choose your video conversion.

175
00:25:31,000 --> 00:25:43,600
And so we have a couple of different profiles. Each user can create
their own profile. I call them full blown but what that basically means
is that

176
00:25:43,666 --> 00:25:56,499
they are going to be converted to all of the different Codec files that
we support. So FLV, MP4, and WebM. But if you don't need all of

177
00:25:56,500 --> 00:25:59,600
those, you can select or deselect the ones that you require.

178
00:25:59,600 --> 00:26:07,366
>> JOHN FOLIOT:
So each end user can create as many sort of conversion pull files as they
want. I mentioned as

179
00:26:07,366 --> 00:26:16,666
well that we support a number of different screen resolutions.   So here
is the 320 by 240 which is the standard 4-3 resolution.

180
00:26:16,666 --> 00:26:26,899
>> JOHN FOLIOT:
I have 480 by 320 which is the 69, 640 by 360 or 640 by 480. We also --
you can also just choose

181
00:26:26,900 --> 00:26:31,900
to not have a video created and just get the timestamp file based on
audio.

182
00:26:32,000 --> 00:26:40,700
>> JOHN FOLIOT:
Again you enter a title and so for the sake of today's demonstration
enter the title of

183
00:26:40,700 --> 00:26:48,633
demonstration. You can if you want provide that short text description
and as most users

184
00:26:48,666 --> 00:26:58,132
>> JOHN FOLIOT:
on campus I choose Cogi at $1.25 a minute here.   And I don't have any
specific vocabulary. So I

185
00:26:58,133 --> 00:27:06,133
didn't choose anything and I click on the submit button and off it goes.

186
00:27:06,133 --> 00:27:12,633
>> JOHN FOLIOT:
One of the things that we noticed however was that some people didn't
actually need

187
00:27:12,633 --> 00:27:20,099
to do the transcript. There was concerns that even at a $1.25 a minute
it was very expensive.

188
00:27:20,100 --> 00:27:26,733
>> JOHN FOLIOT:
And so we provided a means that if you actually have your own flat text
transcript, you can actually

189
00:27:26,733 --> 00:27:29,499
insert the transcript here.

190
00:27:29,500 --> 00:27:39,266
>> JOHN FOLIOT:
So if I go back a couple of slides, you will see right here it says I
already have a transcript. And

191
00:27:39,266 --> 00:27:45,866
so if you click on that, then it changes, it removes the caption
providers and it gives you

192
00:27:45,866 --> 00:27:54,232
>> JOHN FOLIOT:
this upload dialogue where you can upload a flat text file.   The file is
quite simply a flat text file. It is

193
00:27:54,233 --> 00:28:03,199
saved in a .txt format. There is no formatting. As a matter of fact, we
tell people to strip any formatting.

194
00:28:03,266 --> 00:28:12,899
>> JOHN FOLIOT:
So don't give us a Word document or any kind of rtf.   We just want a flat
txt file. UTF coding is fine.

195
00:28:12,900 --> 00:28:21,833
And no line breaks are really necessary. Line breaks will be picked up
or they will be automatically inserted by the system.

196
00:28:21,833 --> 00:28:29,633
>> JOHN FOLIOT:
What we do tell people is do a carriage return at the end of a period.
So if you've got your own

197
00:28:29,633 --> 00:28:40,133
transcript file you can insert it into the system. One of the advantages
there is that right now the cost of this system, the only cost that's
passed on to

198
00:28:40,133 --> 00:28:51,366
the end user is the actual outsourcing to the transcription company.   I
have one faculty member whose students are actually creating videos as
part of sort of their class project or

199
00:28:51,366 --> 00:28:58,432
end of year project and the students themselves are providing the
transcript. They do it themselves.

200
00:28:58,433 --> 00:29:06,599
And so they can insert that in to the system at which point the
timestamping and the alignment and everything else is provided at no
charge.

201
00:29:06,600 --> 00:29:14,966
>> JOHN FOLIOT:
So we listened very carefully to this complaint about the cost of
creating caption files. And so all

202
00:29:14,966 --> 00:29:20,899
of the conversion and the timestamping is mechanically done so it doesn't
cost anything.

203
00:29:20,900 --> 00:29:28,666
>> JOHN FOLIOT:
If I have a transcript file I can insert it there and finally I hit
upload and the file is uploaded in to the

204
00:29:28,666 --> 00:29:32,666
system. You get a little dialogue box and off you go.

205
00:29:32,933 --> 00:29:38,933
>> JOHN FOLIOT:
Depending on your bandwidth and the size of the file, you know, it takes
a minute or three to actually

206
00:29:38,933 --> 00:29:46,933
upload the video in to the system and then you are brought to the next
page in the workflow system.

207
00:29:46,933 --> 00:29:51,333
>> JOHN FOLIOT:
So you will notice that each job is given a unique identifying number and
this particular

208
00:29:51,333 --> 00:30:01,899
demonstration job is 591. That job identification number is also used to
create essentially a work ticket number.

209
00:30:01,900 --> 00:30:08,700
>> JOHN FOLIOT:
It is used in the invoicing process that's done by the transcription
companies. And so it is the

210
00:30:08,700 --> 00:30:14,300
unique identifier of the project so that we can track it through the
system if we have to.

211
00:30:14,300 --> 00:30:20,266
>> JOHN FOLIOT:
You get a little bit of other sort of basic information, the title, the
description, the date that

212
00:30:20,266 --> 00:30:31,399
it was submitted. One of the things that was really important to us was
that we are not -- the system again is a pass through system.

213
00:30:31,400 --> 00:30:39,100
>> JOHN FOLIOT:
We are not at this time in a situation or prepared to do video hosting.
Not that it couldn't be worked

214
00:30:39,100 --> 00:30:46,233
out but that was not our goal when we set the system up. We have had some
conversations with a couple
215
00:30:46,233 --> 00:30:56,799
of people on campus that are looking to have the sort of hosting
situation. But we don't have it at this point.

216
00:30:56,800 --> 00:31:04,433
>> JOHN FOLIOT:
The people at Sacramento asked a question.   Notice the option of data
mining only. And is this

217
00:31:04,433 --> 00:31:13,499
through Docsoft? The answer to that is yes.    The system that we are
using it relies on two servers.

218
00:31:13,500 --> 00:31:20,500
>> JOHN FOLIOT:
One is the Docsoft appliance as well as a custom server that manages all
of the additional business

219
00:31:20,500 --> 00:31:29,300
logic that I am showing you here.   So the data mining we leave it there
simply because it is available.

220
00:31:29,300 --> 00:31:40,433
If anybody has used the system you will know that the speech-to-text,
sort of the mechanical or machine speech-to-text, the quality varies
greatly.

221
00:31:40,433 --> 00:31:48,166
In our early testing we were only getting -- we were getting results as
poor as 20 percent accuracy, 30 percent accuracy which was clearly

222
00:31:48,166 --> 00:31:58,599
not good enough. Somebody asked is this a homegrown system or a
commercial system. This is a homegrown system. This is something

223
00:31:58,600 --> 00:32:07,466
that we put together on campus based on sort
of some understandings and some custom development as well.

224
00:32:07,500 --> 00:32:14,933
>> JOHN FOLIOT:
So getting back to the screen here, you will notice that we have an
expiratory date. So right now
225
00:32:14,933 --> 00:32:21,833
when the videos are converted in to these different file formats. We will
store them on the server for 30 days.

226
00:32:21,833 --> 00:32:29,766
And at the end of 30 days they are automatically removed from the system
simply because we don't have storage space.

227
00:32:29,766 --> 00:32:37,566
>> JOHN FOLIOT:
So you will notice that my files or the conversion status is that the
conversion to these different file

228
00:32:37,566 --> 00:32:43,699
formats is in process as is the transcription status.

229
00:32:43,700 --> 00:32:55,000
And so what happens is that the transcription status will always be in
process even though -- so the first step is that we need to do the
conversion.

230
00:32:55,000 --> 00:33:02,733
>> JOHN FOLIOT:
And we can't go any further in the process until we have at the very
least the MP3. So the

231
00:33:02,733 --> 00:33:09,799
system will convert the file to an MP3 first as you can see it was done
fairly quickly.

232
00:33:09,900 --> 00:33:16,466
>> JOHN FOLIOT:
And so at that point the end user can download the MP3.   But again if you
remember on one of

233
00:33:16,466 --> 00:33:22,232
the earlier slides we have some internal logic in the system that sends
an e-mail off

234
00:33:22,233 --> 00:33:31,433
to the transcription company that was selected. And it says that the MP3
file is ready to be picked up and converted in to text.

235
00:33:31,433 --> 00:33:40,966
So the transcription status will be in process until such time as the
transcription company reloads the flat text file in to the system.

236
00:33:40,966 --> 00:33:47,966
>> JOHN FOLIOT:
If you have your own text file it will start to happen once the
conversion of these file formats is done.

237
00:33:47,966 --> 00:33:58,266
But if it is being outsourced and you have chosen 48 hours, then it will
be sort of in process for 48 hours until the file comes back.

238
00:33:58,333 --> 00:34:05,799
>> JOHN FOLIOT:
In this particular example I had a text file all ready.   And so obviously
it is ready to be downloaded. I

239
00:34:05,800 --> 00:34:13,100
mean it is kind of silly in this particular instance because I uploaded
the text file and it is ready for download.

240
00:34:13,100 --> 00:34:21,633
>> JOHN FOLIOT:
We didn't bother worrying about the skewed logic there but it has not yet
been timestamped. And

241
00:34:21,633 --> 00:34:32,166
the timestamping again is dependent on the MP3 file because the Docsoft
appliance would actually apply as the timestamping.

242
00:34:32,166 --> 00:34:38,066
>> JOHN FOLIOT:
So a couple of minutes later and you can see the completion.   It took me
roughly 18 minutes to

243
00:34:38,066 --> 00:34:44,599
convert the video in to the different file formats.   And so they are all
available for download.

244
00:34:44,700 --> 00:34:52,233
>> JOHN FOLIOT:
Once all of the video files have been converted to their sort of web
ready format, the system sends
245
00:34:52,233 --> 00:34:59,633
an e-mail to the individual user informing them that the videos are, in
fact, ready to be picked up.

246
00:34:59,866 --> 00:35:09,466
As I said at this point the most mechanical part of the system is users
log in to their system, they go to the project page and they have to

247
00:35:09,466 --> 00:35:18,266
download the files on to their local drive so that they can be uploaded
to their web server or put in their media asset management tool.

248
00:35:18,266 --> 00:35:24,032
>> JOHN FOLIOT:
We are hoping in the future to be able to have rather than make it
available for download that it

249
00:35:24,033 --> 00:35:33,166
could just be fired off to a storage facility, but there is a couple of
issues about FTP file permissions and whatnot. So we need to work those
out.

250
00:35:33,166 --> 00:35:40,466
>> JOHN FOLIOT:
But at any rate the file, they receive an e-mail saying that the files
are ready to download.

251
00:35:40,466 --> 00:35:49,232
>> JOHN FOLIOT:
If they have chosen the transcription process that's going to take more
than 24 hours, one of the

252
00:35:49,233 --> 00:35:54,099
things that we heard was that oftentimes they need to get the video up as
soon as possible.

253
00:35:54,200 --> 00:36:03,200
And so I would rather not discourage people from putting up their videos.
I don't want the captions to be the roadblock.

254
00:36:03,200 --> 00:36:09,266
>> JOHN FOLIOT:
Obviously I would love to have, you know, the captions go live with the
video right away but in
255
00:36:09,266 --> 00:36:13,266
some instances I am prepared to accept a little bit of water with the
wine.

256
00:36:13,266 --> 00:36:21,399
>> JOHN FOLIOT:
And if the captions catch up a day or two later then that's what it is.
The e-mails are

257
00:36:21,433 --> 00:36:28,399
actually sent individually. So a second e-mail is also sent out when the
caption files are ready to be downloaded as well.

258
00:36:28,400 --> 00:36:40,666
The other thing that happens is that once the process is done, the
project file creates a little sort of preview window here.

259
00:36:40,666 --> 00:36:48,466
>> JOHN FOLIOT:
Anybody that's using video probably recognizes the JW FLV player which is
the player that

260
00:36:48,466 --> 00:36:58,866
we are using. And so it is embedded in the page there. So that you can
review the file, you can see the caption associated with the video.

261
00:36:58,866 --> 00:37:10,866
One of the things that has been occurring occasionally is that once the
timestamp files are created people notice that they are off a little bit
or

262
00:37:10,866 --> 00:37:16,232
>> JOHN FOLIOT:
in the transcription status if it was outsourced, there may be some
spelling mistakes in there.

263
00:37:16,233 --> 00:37:23,633
If that's the case, they can take the flat text file, they can do minor
edits.

264
00:37:23,633 --> 00:37:31,966
And you recall that in the system you can say I have a text transcript
already. And so they can reupload it and it would be retimestamped.
265
00:37:31,966 --> 00:37:42,066
>> JOHN FOLIOT:
So error correction like that. The dfxp file and the subrip file are
also essentially flat text files.

266
00:37:42,100 --> 00:37:54,500
So if you need to do some minor timestamp tweaking you can open that up
in the text file. You have the ability in process to do some fine-
tuning.

267
00:37:54,500 --> 00:38:04,500
Finally at the bottom of each project page we provide a link to download
the JW FLV player. We bought a license for it.

268
00:38:04,500 --> 00:38:12,766
The license is relatively cheap. I think Sean paid $50, $60 for the
license. We did a little bit of additional branding to it.

269
00:38:12,766 --> 00:38:25,899
The most recent version of the JW player, the module for displaying
captions in the player is a third party sort of stand-alone module. What

270
00:38:25,900 --> 00:38:34,166
we did is we actually took the module and we integrated it in to the
player and then rebundled it. The player, of course, is done in Flash.

271
00:38:34,166 --> 00:38:40,232
>> JOHN FOLIOT:
And so we open it up in to a Flash editor and made a minor tweak there so
that the player that

272
00:38:40,233 --> 00:38:48,366
we offer for download has the module built in.   We also provide some copy
and paste code here.

273
00:38:48,366 --> 00:38:57,432
Both raw sort of HTML code .. we had a couple people using content
management systems, blogging tools and whatnot.

274
00:38:57,433 --> 00:39:06,766
And so we have a base URL so you specify where the files are located and
then we generate some copy and paste code there.
275
00:39:06,766 --> 00:39:14,666
So that people can add it to their either -- in to their Web page or in
to their blog pages or what have you.

276
00:39:14,666 --> 00:39:19,632
>> JOHN FOLIOT:
You notice perhaps along the top that there are a couple of additional
buttons. So under the library

277
00:39:19,633 --> 00:39:27,033
button you have a media library and each of the projects has an entry
that looks like this.

278
00:39:27,033 --> 00:39:36,866
Gives you a screen capture of the first frame, which we also make
available as a download right here, the preview image.

279
00:39:36,866 --> 00:39:46,899
>> JOHN FOLIOT:
So you have got a screen capture that's available.   I'm sorry.   Got to
get back to the right slide. You

280
00:39:46,900 --> 00:39:54,133
can -- if you have a number of projects you can sort by submission date,
title or duration. And you have some basic details.

281
00:39:54,133 --> 00:40:02,899
>> JOHN FOLIOT:
Clicking on the details link takes you back to the, sorry, takes you back
to the actual project page

282
00:40:02,900 --> 00:40:08,633
there. And so you have a number of projects depending on how you go.

283
00:40:08,633 --> 00:40:16,099
As I said each user has their own unique user account.   And so the user
account, user
preferences look like that.

284
00:40:16,100 --> 00:40:24,933
>> JOHN FOLIOT:
So if you need you can go in and change passwords.   I mention that all
throughout the
285
00:40:24,933 --> 00:40:33,399
process e-mails are sent to the end users. If you don't want to receive
all of those e-mails, you can choose not to get those notifications.

286
00:40:33,466 --> 00:40:41,266
>> JOHN FOLIOT:
We also have a couple of settings that can be modified.   The
accessibility settings ... when we

287
00:40:41,266 --> 00:40:45,832
set up the system and the system is now almost three years old.

288
00:40:45,833 --> 00:40:51,633
>> JOHN FOLIOT:
We had one person complaining about sort of some of the JavaScript that
was being used in the

289
00:40:51,666 --> 00:40:59,499
interface and they didn't want to have it.   You can disable the scripting
part of the system.

290
00:40:59,500 --> 00:41:08,700
You just get more page scrolling and it takes longer to go through the
workflow. But it
really doesn't have a lot of practical impact.

291
00:41:08,700 --> 00:41:18,366
But the other thing that I mention is that we have a conversion workflow
profile. And so each user can create as many profiles as they want.

292
00:41:18,366 --> 00:41:25,999
You just give it a number. And so as I mentioned earlier I called it
full blown but you can call it whatever you want.

293
00:41:26,000 --> 00:41:32,800
>> JOHN FOLIOT:
You get to choose the final output size as well as the formats that are
going to be converted in.

294
00:41:32,800 --> 00:41:40,166
>> JOHN FOLIOT:
So, for example, if you know that you really only want an FLV file, then
that's all you need to
295
00:41:40,166 --> 00:41:46,766
convert to. It does speed up the process.   And it also saves us a little
bit of CPU.

296
00:41:46,766 --> 00:41:53,932
One of the attendees asks who is paying for the captioning.   The
individual user pays for their own captioning.

297
00:41:53,933 --> 00:42:01,966
>> JOHN FOLIOT:
So it would be the department or the faculty member is directly billed.
And so as I

298
00:42:01,966 --> 00:42:07,432
mentioned we have .. when you set up the user profile we capture billing
information

299
00:42:07,433 --> 00:42:15,966
>> JOHN FOLIOT:
so that the transcript company sends the invoice directly to the end
user. So this is the conversion

300
00:42:15,966 --> 00:42:24,266
workflow profile so that you can go in and create as many or as few as
you want.

301
00:42:24,266 --> 00:42:35,499
We also have now .. because I am in administrator I am giving you a
screen shot of the administrator panel. And so most users will not

302
00:42:35,500 --> 00:42:41,766
>> JOHN FOLIOT:
But the system administrator, so Sean and I do.   So there is a number of
back end things that we

303
00:42:41,766 --> 00:42:47,899
can do to set up the system. So one of the first important things is
managing the department.

304
00:42:47,900 --> 00:42:54,433
>> JOHN FOLIOT:
So kind of related to that question of who pays for the captioning, the
department does. So
305
00:42:54,433 --> 00:43:02,433
we have a listing of all the different, you know, here is one that Sean
set up just as a test account, the Acme labs.

306
00:43:02,433 --> 00:43:07,133
>> JOHN FOLIOT:
So we capture a contact name, a person that we are actually dealing with
in the department and we

307
00:43:07,133 --> 00:43:14,166
have their e-mail address. And these three fields here are obligatory
when creating a user profile.

308
00:43:14,166 --> 00:43:18,332
>> JOHN FOLIOT:
We need an e-mail address and need a contact name and we need a
department. We have a

309
00:43:18,366 --> 00:43:24,132
contact phone number so if we need to get in touch with the people we
can. We have the mailing

310
00:43:24,133 --> 00:43:31,633
address so again when the transcript company creates the invoice they
just send the invoice straight to the user.

311
00:43:31,633 --> 00:43:38,333
>> JOHN FOLIOT:
So for Sean and I, again, it is as hands off as we can make it.   You can
in fact edit that information.

312
00:43:38,333 --> 00:43:46,899
And so .. you know .. I dont' have the full screen to show you. But
moving along the mailing address, the e-mail or web address if we have,
and then

313
00:43:46,900 --> 00:43:52,633
you can edit or delete or when you are editing you get this gray look.

314
00:43:52,633 --> 00:43:59,266
>> JOHN FOLIOT:
So we can change the contact name, e-mail address, the mailing address,
et cetera, et cetera.
315
00:43:59,266 --> 00:44:04,332
>> JOHN FOLIOT:
We also have a similar system for the different transcript services.   So
right now I mention we are

316
00:44:04,333 --> 00:44:10,199
dealing with while the Stanford transcription is a test account, it
doesn't really do anything.

317
00:44:10,200 --> 00:44:15,500
>> JOHN FOLIOT:
But we are dealing with four transcript companies that have given us
different price points.

318
00:44:15,533 --> 00:44:20,299
>> JOHN FOLIOT:
We can add as few or as many as we want.   So one of the things that I was
concerned

319
00:44:20,300 --> 00:44:25,833
about moving forward is that I didn't want to put all of our eggs in one
basket.

320
00:44:25,833 --> 00:44:34,466
So as the system continues to grow if we need to add other transcription
companies or if other transcription companies come forward that can

321
00:44:34,466 --> 00:44:44,499
give us a better price we can add a new transcript company in to the
service offering and so again our users have a choice.

322
00:44:44,500 --> 00:44:51,433
>> JOHN FOLIOT:
So the system in many ways is almost like a brokerage or a user bazaar
and we give

323
00:44:51,433 --> 00:44:56,666
users the options, but they actually engage directly with the transcript
company.

324
00:44:56,666 --> 00:45:06,032
>> JOHN FOLIOT:
Our system simply works as the brokerage and manages it all sort of
behind the scenes. We can
325
00:45:06,033 --> 00:45:17,133
also manage our users directly.   And so over here we can decide whether
the users are billable or not.

326
00:45:17,133 --> 00:45:23,599
>> JOHN FOLIOT:
So a couple of the systems we had a couple of people come to us that
wanted to use the

327
00:45:23,600 --> 00:45:33,100
system. We are always going to have the transcript. So, for example,
students and so we can set up the system that a student could log in.

328
00:45:33,100 --> 00:45:38,433
>> JOHN FOLIOT:
They could upload a file but they do not have access to the transcript
companies and that's

329
00:45:38,433 --> 00:45:47,466
kind of something that we added after the system was initially created
because we had some demand for that.

330
00:45:47,466 --> 00:45:52,266
>> JOHN FOLIOT:
We didn't want to be in a situation where we were trying to capture
billing information from

331
00:45:52,266 --> 00:45:58,999
students because as you know students come and go and if they skipped on
the bill we would be left holding the bag.

332
00:45:59,000 --> 00:46:04,600
>> JOHN FOLIOT:
So we have created a system where if you have got a file you can still
get the timestamping and

333
00:46:04,600 --> 00:46:15,200
the conversion done but you do not have access to the billable part of
the system. As well for security reasons

334
00:46:15,200 --> 00:46:23,733
if you try and log on and you fail three times you are locked out. At
which point we receive an e-mail saying that a particular user was locked
out.

335
00:46:23,733 --> 00:46:28,366
>> JOHN FOLIOT:
And we will contact them and ask if they were having problems, or was
somebody

336
00:46:28,366 --> 00:46:35,432
trying to hack their account and so then we can reset it based on the
answers to those questions.

337
00:46:35,433 --> 00:46:44,699
>> JOHN FOLIOT:
We also as I said when somebody first signs up we receive an e-mail, you
know, requesting to

338
00:46:44,700 --> 00:46:51,700
have an account with the system. And so for the time being Sean and I
have set it up that it is a manual approval process.

339
00:46:51,700 --> 00:46:57,066
>> JOHN FOLIOT:
We could automate that if we wanted to but again we wanted to make sure
that the people that are

340
00:46:57,066 --> 00:47:03,299
using the system were, in fact, affiliated with the university.

341
00:47:03,366 --> 00:47:10,799
>> JOHN FOLIOT:
As far as the pricing is concerned you select -- once you have created a
transcript service you

342
00:47:10,800 --> 00:47:18,766
select the name of the transcript service and then you can add as many
different price services as possible.

343
00:47:18,766 --> 00:47:24,232
>> JOHN FOLIOT:
Here the price per minute and we -- excuse me, we create a description
here and the

344
00:47:24,233 --> 00:47:31,799
description is actually what is read here. So we actually write out two
business days project read on at the price point.

345
00:47:31,800 --> 00:47:41,633
That's what is displayed on the video upload page to the end user. So we
couldn't find a more elegant way than that. So it is a little kludgy but
at

346
00:47:41,633 --> 00:47:52,699
least it is accurate. So that's basically the system from end to end.   As
I said we provide a copy and paste of the

347
00:47:52,700 --> 00:48:03,900
HTML code. One of the things that we are looking at is integrating it
with Drupal. On our campus Drupal has become a very

348
00:48:03,900 --> 00:48:11,300
popular content management system. My guess is within the next two to
three years probably about 80 percent of the web content

349
00:48:11,300 --> 00:48:22,266
on Stanford's campus will be delivered via the Drupal system. So we are
looking at having a tighter integration with the Drupal system.

350
00:48:22,266 --> 00:48:30,666
>> JOHN FOLIOT:
The -- we also have tutorials on captioning Youtube videos as well as
support for iOS.

351
00:48:30,666 --> 00:48:37,666
Somebody asked a question "Is the audio mine option free of cost to the
users?"

352
00:48:37,666 --> 00:48:46,499
>> JOHN FOLIOT:
It is, it is free of cost. The problem is that the accuracy .. it's just
.. I mean if you have a really

353
00:48:46,500 --> 00:48:57,333
sort of good speaker with no accent, you know somebody from the Midwest
and that has been well mic'd and you have really good audio quality

354
00:48:57,366 --> 00:49:00,932
then the accuracy is not bad.

355
00:49:00,933 --> 00:49:07,033
>> JOHN FOLIOT:
By our experience, however, having that kind audio quality going in is
something of a pipe

356
00:49:07,033 --> 00:49:15,533
dream more often than not. Its you know, a boom mic at the back of the
room or you've got

357
00:49:15,533 --> 00:49:23,666
>> JOHN FOLIOT:
faculty members or presenters with thick and heavy accents. At that point
the quality of the

358
00:49:23,666 --> 00:49:30,732
speech-to-text goes way down and it really is a question of the law of
diminishing returns.

359
00:49:30,733 --> 00:49:35,166
>> JOHN FOLIOT:
When you get back 30 percent accuracy you might as well get back 0
percent accuracy

360
00:49:35,166 --> 00:49:40,832
because it is going to take you so long to try and clean that up that
it's really not worthwhile.

361
00:49:40,833 --> 00:49:47,799
>> JOHN FOLIOT:
However if people want to do that and then use that as a basis to clean
it up and provide a text

362
00:49:47,800 --> 00:49:54,333
transcript, and then insert that transcript in to the system, they are
free to do so and there is no cost.

363
00:49:54,333 --> 00:49:59,399
>> JOHN FOLIOT:
We will still do the Codec conversion and create the sort of web ready
video file. We will still

364
00:49:59,400 --> 00:50:06,900
generate the timestamp files. You get the copy and paste code. So all of
the other pieces of creating sort of a captioned video for web delivery

365
00:50:06,900 --> 00:50:14,700
will still be delivered by the system. So the -- another question

366
00:50:14,700 --> 00:50:21,800
is when a vender is chosen who pays for it. The professor, the
department, disability services, et cetera, the client.

367
00:50:21,800 --> 00:50:29,633
So -- when -- I mean I hate to use business terms on campus sometimes
because it bristles some people's necks.

368
00:50:29,633 --> 00:50:39,566
But this really is a business relationship that we've set up .. between
somone who's creating video content and the transcription company.

369
00:50:39,600 --> 00:50:45,200
>> JOHN FOLIOT:
And so all we do is we create a business relationship for them.    If it is
disability services that

370
00:50:45,233 --> 00:50:52,566
wants to set up an account, then they can do so. And any videos that
they set send in to the system gets billed back to them.

371
00:50:52,566 --> 00:50:58,499
>> JOHN FOLIOT:
So, of course, Sean who works in Stanford's office of accessible
education which is our

372
00:50:58,500 --> 00:51:04,766
disability services department, when he uses the system, the bill goes to
them.

373
00:51:04,766 --> 00:51:11,966
>> JOHN FOLIOT:
But one of the things that we have done is that we have tried to push the
costs back to the end user

374
00:51:11,966 --> 00:51:17,532
not in a negative way but by going around and explaining sort of the
value of having captioned files.
375
00:51:17,533 --> 00:51:24,466
>> JOHN FOLIOT:
And so one of the groups on campus that I am working with is it is called
tech training and as

376
00:51:24,466 --> 00:51:29,332
part of education group that's located in IT services.

377
00:51:29,333 --> 00:51:36,433
>> JOHN FOLIOT:
And every Friday afternoon on campus they have a 90-minute free kind of
presentation where they

378
00:51:36,433 --> 00:51:42,066
bring in people to talk about various technical subjects.     I have done
presentations there.

379
00:51:42,133 --> 00:51:48,333
They have brought in third party venders from, you know, Adobe and
Microsoft and they do these 90 -minute

380
00:51:48,400 --> 00:51:58,433
sessions on - you know - a myriad of different topics. They are
videotaping all these sessions to create their own sort of video library
and they

381
00:51:58,433 --> 00:52:06,266
are using the captioning system. And so they were set up as a client in
the system and they are using

382
00:52:06,266 --> 00:52:17,566
Cogi right now. So when they upload the video, Cogi does the speech-to-
text and fires back the flat text file and they send the invoice straight
to

383
00:52:17,566 --> 00:52:26,899
the training group at IT services and they pay it directly.    So, you
know, it really is set up as a client process like that.

384
00:52:26,900 --> 00:52:31,466
>> JOHN FOLIOT:
Somebody asked do we have a policy on campus requiring professors to
caption videos. We do
385
00:52:31,466 --> 00:52:41,766
not have a policy like that at this time.   And so that's the answer. I
don't have a policy at this time.

386
00:52:41,766 --> 00:52:51,399
We are trying really hard to sort of sell the benefits of captions in
that it makes the videos more searchable, there is a benefit to end users

387
00:52:51,400 --> 00:52:59,533
>> JOHN FOLIOT:
and to students as well, especially students that are, you know, English
is not their first language.

388
00:52:59,566 --> 00:53:09,299
But we don't -- I don't have a stick right now. So I am walking around
with lots of carrots. One of the things that we did -- so I mentioned
integration

389
00:53:09,300 --> 00:53:19,966
with Drupal. So we have created a custom input type for those that know
Drupal using CCK .. the video can be hosted locally on the same server as

390
00:53:19,966 --> 00:53:30,432
the drupal instance or remotely and so.. one of the things that Stanford
has on campus is .. .. there is a group called Stanford Video and they

391
00:53:30,433 --> 00:53:42,433
maintain a streaming video server as opposed to pseudo streaming and you
can host up to a gigabyte of, no, excuse me, a 100 gigabytes of

392
00:53:42,433 --> 00:53:47,766
>> JOHN FOLIOT:
So for smaller departments or faculty members they can set up an account
at Stanford video and

393
00:53:47,766 --> 00:53:55,999
they can move their videos and host them there and they get true video
streaming as opposed to pseudo streaming.

394
00:53:56,000 --> 00:54:03,700
>> JOHN FOLIOT:
And so unfortunately I can't really show you some examples, but one of
the other things that the
395
00:54:03,700 --> 00:54:10,866
caption files that are created can also be used in third party
applications like Youtube.

396
00:54:10,866 --> 00:54:16,266
>> JOHN FOLIOT:
So if anybody is using Youtube and I know a lot of faculty members and
students use Youtube as

397
00:54:16,266 --> 00:54:31,499
kind of the final resting place for their videos, the caption files that
are being created can also be uploaded to Youtube and Youtube will then
use

398
00:54:31,500 --> 00:54:40,033
them in their player. And so one of the nice things about the Youtube
system is that you can search for a particular line in the lecture and
you can click

399
00:54:40,033 --> 00:54:43,599
on that line and it will take you to that place in
the video.

400
00:54:43,600 --> 00:54:50,100
>> JOHN FOLIOT:
So again, when we go around and talk to people and kind of explain the
benefits of having a time

401
00:54:50,100 --> 00:54:56,166
synchronized caption file with the video we certainly point to that as
being a useful thing.

402
00:54:56,166 --> 00:55:02,466
>> JOHN FOLIOT:
So there is a couple of other questions here.   One person says how many
people are involved

403
00:55:02,466 --> 00:55:11,932
in the selling. I am not really sure what you mean by that question. The
selling is done sort of automatically.

404
00:55:11,933 --> 00:55:19,799
We created the relationship with the transcription company by adding them
in to the system as a vender.

405
00:55:19,800 --> 00:55:29,966
And then the sort of offer of work is the e-mail that gets sent to the
transcription company when somebody uploads the video. So that's the
offer

406
00:55:29,966 --> 00:55:39,166
of work. The e-mail that they receive has that job number that I
indicated. Has a location where
the MP3 file is and says that, you know, this

407
00:55:39,166 --> 00:55:47,399
particular department, this particular client and we give them all the
billing information has this file that needs to be transcribed and here
is

408
00:55:47,400 --> 00:55:51,733
the time frame and you agree to this type of price.   Please do the job.

409
00:55:51,733 --> 00:55:59,233
>> JOHN FOLIOT:
Wow there is a lot of questions sort of coming right now, so let me just
finish through. I only have

410
00:55:59,233 --> 00:56:06,566
a couple more slides. And then we can go back and I will try and process
some of these questions. We also have a Q and A session at the

411
00:56:06,566 --> 00:56:20,299
end of the larger seminar. So a couple of other
systems here, so iTunes and iOS platform, the SCC file format is really
tricky to work with

412
00:56:20,300 --> 00:56:31,566
but there is some third party tools now that allows for subtitling and we
also use it for captioning. And there is a MacIntosh tool that is free
to use

413
00:56:31,566 --> 00:56:41,666
that will do that. So we don't get too twitchy about whether it is a
subtitle or a closed caption. I mean what's important is that the text
is being provided
414
00:56:41,666 --> 00:56:48,166
>> JOHN FOLIOT:
to those that can't hear. So, you know, these are some of the ways that
we provide for the subtitling

415
00:56:48,166 --> 00:57:00,766
for Mac OS. So the tool is iSubtitle and here is the URL there. It kind
of side steps having to deal with the SCC file format which required a
lot of

416
00:57:00,766 --> 00:57:11,199
work earlier on. A couple of screen shots here. I am not going to spend
a lot of time on that. Capturing for the Android platform it is still
tricky, it

417
00:57:11,200 --> 00:57:23,200
is immature but it requires the SRT file format which is what we are
generating right now. So it is doable but it is still kind of awkward.

418
00:57:23,200 --> 00:57:30,500
>> JOHN FOLIOT:
And so there is some additional resources.   Some URLs here.   I know
Howard will be

419
00:57:30,600 --> 00:57:34,966
providing all of this information at the end of the seminar.

420
00:57:34,966 --> 00:57:43,266
>> JOHN FOLIOT:
So if you need to get any of these URLs you can.   And with that there is
my contact information

421
00:57:43,266 --> 00:57:52,066
there. And I am happy to answer questions if you want to send me an e-
mail to follow up directly I can.

422
00:57:52,066 --> 00:58:08,299
I am looking at the time now. Howard, what is the process? Are we
taking a break now or is it question and answer for the next 15 minutes?

423
00:58:08,300 --> 00:58:15,400
>> JOHN FOLIOT:
Howard says we are going to do some question and answer for the next 15
minutes. So a number
424
00:58:15,400 --> 00:58:18,866
of people have typed some questions in to the chat section.

425
00:58:18,866 --> 00:58:27,666
I am not going to use names here. But somebody asked how can I create
the HTML files without knowing where the video will be hosted.

426
00:58:27,666 --> 00:58:41,232
That's a good question. It really depends on how you are going to be
using the video file. I am looking for the -- I have to go back some
more.

427
00:58:41,266 --> 00:58:52,599
>> JOHN FOLIOT:
Sorry, I am looking for the particular slide.   So I can show you.   There
we go here. So what

428
00:58:52,600 --> 00:59:03,966
happens is that in this, this HTML code here the files that are created
and we have a tutorial that's associated with this website as well.

429
00:59:03,966 --> 00:59:16,766
So the files that are created always have the same name.   So it is always
media.flv, media.webm, media.mp4 and then caption.srt,

430
00:59:16,766 --> 00:59:21,066
caption.xml, et cetera, et cetera. And so

431
00:59:21,066 --> 00:59:30,099
what we tell people to do is to download all those files and put them in
a directory and name that directory or the file folder the name of the

432
00:59:30,100 --> 00:59:37,500
>> JOHN FOLIOT:
project. So, for example, in this particular project I used the name of
demonstration. So the tutorial

433
00:59:37,500 --> 00:59:44,800
says to download all those files and create a directory called
demonstration and then FTP that on to the server.

434
00:59:44,800 --> 00:59:52,600
>> JOHN FOLIOT:
And so this code here because the name of the file or the name of the
project was called

435
00:59:52,600 --> 01:00:01,100
demonstration, there is a little bit of dynamic writing here that
actually paths you to that particular directory.

436
01:00:01,100 --> 01:00:07,066
>> JOHN FOLIOT:
So if you put the demonstration directory at the root of your web
location, this copy and

437
01:00:07,066 --> 01:00:10,499
paste code just works because it has been figured out that way.

438
01:00:10,500 --> 01:00:18,233
>> JOHN FOLIOT:
So we have a sort of FAQ file that's associated with the website that
explains how to do that. In

439
01:00:18,233 --> 01:00:29,799
the case of a blogging platform, you have to upload your video and
transcript files to the final location before you can use the copy and
paste

440
01:00:29,800 --> 01:00:40,933
code. And you type in the location of where that stuff is. Click on the
refresh button and it dynamically rewrites this code here so that it will

441
01:00:40,933 --> 01:00:52,199
point to the final location. So I don't know if that answers that
particular question. If it is not clear, you can certainly send me an e-
mail and I can

442
01:00:52,200 --> 01:01:04,233
follow up some more. Somebody from West Texas A&M ask what product is
used to match the flat text file to the video. So again as I mention we

443
01:01:04,233 --> 01:01:08,199
have two servers that are working in tandem.

444
01:01:08,200 --> 01:01:20,600
One is the Docsoft appliance from Docsoft. And I believe it is
docsoft.com and it is the tool that actually does the timestamping.

445
01:01:20,600 --> 01:01:28,233
>> JOHN FOLIOT:
And so the -- yeah. That's the tool that does that.   The second server
is one that we had custom

446
01:01:28,233 --> 01:01:40,033
created. Docsoft did the work for us and it is the one that manages all
the business logic. And it also does all of the Codec conversion.

447
01:01:40,033 --> 01:01:58,566
The next question is --. Somebody asked a question something about a --
explain the benefits to faculty, I am not really under --

448
01:01:58,566 --> 01:02:06,366
>> JOHN FOLIOT:
so in terms of explaining the benefits to faculty, yeah, so anybody
that's involved with accessibility

449
01:02:06,366 --> 01:02:12,932
issues or disability resource issues, you know, yeah, I buy lots of
coffee.

450
01:02:12,933 --> 01:02:20,233
>> JOHN FOLIOT:
What can I tell you? A lot of it is kind of one-on-one.   So evangelizing
and meeting with people,

451
01:02:20,266 --> 01:02:29,332
we have a website that we work on. I also do presentations.   I mention I
have done tech briefings on captioning.

452
01:02:29,333 --> 01:02:34,399
>> JOHN FOLIOT:
The other thing is that both Sean and I have a really good relationship
with a lot of the

453
01:02:34,400 --> 01:02:42,633
technology people on campus, a lot of the web masters and so through them
we will arrange meetings with various faculty members. Or we'll

454
01:02:42,633 --> 01:02:47,199
sit down, we have eight different sort of presentations that we do.

455
01:02:47,200 --> 01:02:52,800
>> JOHN FOLIOT:
But we have some stock speaking points and some things that we show
people. So I didn't

456
01:02:52,800 --> 01:03:00,766
mention but I showed you in Youtube that screen shot where you can search
inside the video to that particular point.

457
01:03:00,766 --> 01:03:10,099
Another thing that Youtube has that's still experimental which is very
cool is the automatic language translation. And so not only are we

458
01:03:10,100 --> 01:03:18,766
getting captions but you get subtitles.
And so we do little demonstrations of that to sort of show how it
benefits everyone.

459
01:03:18,766 --> 01:03:24,466
>> JOHN FOLIOT:
We lean real hard in to the searchability of videos and the fact that
caption files are being indexed

460
01:03:24,466 --> 01:03:31,866
by Youtube which is Google. There is an SCO benefits search engine of
doing that. Optimization benefit of doing that.

461
01:03:31,866 --> 01:03:42,399
So, yeah, but a lot of it is, you know, based on working, personal
relationships and just
working with people to make them understand.

462
01:03:42,400 --> 01:03:54,266
>> JOHN FOLIOT:
Somebody says are you going to speak to what this is written in?    So we
worked with Docsoft.

463
01:03:54,266 --> 01:03:58,532
We actually outsource the production of this entire system to them.

464
01:03:58,533 --> 01:04:03,133
>> JOHN FOLIOT:
We bought the Docsoft appliance and then we had some custom work done.
They are a

465
01:04:03,133 --> 01:04:15,033
Windows shop. The Docsoft shop appliance is running on ISS and all the
other logic is written in asp or aspx I believe. It is all written to
work on a

466
01:04:15,033 --> 01:04:24,033
Windows platform. I would love, I would love to find somebody who would
be interested in taking some of the business logic, I mean the Docsoft

467
01:04:24,033 --> 01:04:31,866
appliance runs on IAS and so there is nothing we can do about it. But
the second server that manages all the business logic and what not..

468
01:04:31,866 --> 01:04:35,032
>> JOHN FOLIOT:
Now that we have got this up and running I would love to find somebody
that .. who would be

469
01:04:35,033 --> 01:04:43,833
interested in taking on the project and doing it in PHP.   I thought I had
a student in the
computer sciences department who was

470
01:04:43,833 --> 01:04:49,966
interested in the project but as students are, he kind of moved on and I
never heard anything more.

471
01:04:49,966 --> 01:04:55,399
>> JOHN FOLIOT:
But I'll tell you honestly, that if I could find someone to do it in PHP
I would open souce it , I would put

472
01:04:55,400 --> 01:05:01,800
it out as free. You still need the Docsoft appliance but all the other
business logic I would give away simply because we need it, right? So I
don't know

473
01:05:01,800 --> 01:05:09,266
 if that answers that question.   But there you go.

474
01:05:09,266 --> 01:05:16,199
Somebody asked the question is this work used for commercial videos that
are shown in classrooms.

475
01:05:16,200 --> 01:05:25,500
>> JOHN FOLIOT:
I am not aware of anybody doing that yet.   There is no reason why you
couldn't. My program is

476
01:05:25,500 --> 01:05:35,000
really focused on online accessibility. So, you know, web delivery.      So
it was really sort of created to address that specific need.

477
01:05:35,000 --> 01:05:44,600
>> JOHN FOLIOT:
But I mean the system of producing timestamped files that can be then
associated to a video if you

478
01:05:44,600 --> 01:05:50,133
had say an old movie that for whatever reason a faculty member was using
a

479
01:05:50,133 --> 01:05:56,466
>> JOHN FOLIOT:
film from the '30s or '40s say that didn't have closed captions, it would
be a little bit of

480
01:05:56,466 --> 01:06:09,466
additional post production that disability services would have to do but
in theory they could

481
01:06:09,466 --> 01:06:17,832
and then it would do the timestamping and create an SCC file that you can
take into final cut pro or somthing and export that video and burn it as

482
01:06:17,833 --> 01:06:20,233
a DVD or whatnot.

483
01:06:20,233 --> 01:06:30,933
The capability of doing that exists but that’s not what the system was
designed to do. I hope that answers that question. Somebody asked what

484
01:06:30,933 --> 01:06:41,166
pots of money do users tap in to to pay for the transcription costs
typically. There you go. That’s the sticking point. And so it
generally comes out

485
01:06:41,166 --> 01:06:54,999
of operational budgets, which is a very euphemism for where they can find
it. That’s been our big stumbling block. When we launched

486
01:06:55,000 --> 01:07:05,033
the system two and a half, three years ago it was right around the time
the local or even the national economy took that massive hit. So right
at the

487
01:07:05,033 --> 01:07:15,333
time when all excised budgets dried up completely. So one of the things
that we have also done is we have been talking to sort of

488
01:07:15,333 --> 01:07:23,699
administrative people in the various departments about, you know, when
they are working on their annual budgets that they have to do every year.

489
01:07:23,700 --> 01:07:33,800
And at my university there are about 60 to 70 percent through that
process now and usually starts after Christmas break and the real
decision

490
01:07:33,800 --> 01:07:44,133
making happens in July or August. We talked to them about thinking about
how much video they are going to be producing. And sort of putting that

491
01:07:44,133 --> 01:07:54,666
-- the cost of transcription in to their working budgets. A lot of the
videos that we are seeing go through the system right now tend to be of
the

492
01:07:54,666 --> 01:08:06,499
short form video file, format. So very much geared to the Youtube
delivery. Videos are usually between three and five minutes. At $1.25

493
01:08:06,500 --> 01:08:21,266
a five minute video is going to cost us $7.50. You can’t buy lunch on
campus for that. Right now it comes from wherever assets can be found.
But

494
01:08:21,266 --> 01:08:30,699
that’s our big challenge right now is getting the administrative people
and the budgeting people to work this in to their average or their basic

495
01:08:30,700 --> 01:08:44,533
yearly budgets. Question about copyright, what if the original videos are
commercial and not personal. So again as I said the system was not

496
01:08:44,533 --> 01:08:57,633
really designed for, you know, creating closed captions for commercial
product. It was created for videos that are being created on campus by

497
01:08:57,633 --> 01:09:09,299
faculty students and staff to be used via web delivery. The system can
be used for commercial videos by the disability resource center. But,
you

498
01:09:09,300 --> 01:09:21,133
know, they are governed by the same copyright rules that exist for, you
know, using any other system. So, you know, this is just an alternative
to

499
01:09:21,133 --> 01:09:31,466
some of the commercial systems that are out there. You know, if you are
a disability resource center that has to do a large amount of

500
01:09:31,466 --> 01:09:42,799
commercial videos, I would say at that point this system probably is not
the right tool for that particular job. This was very much sort of

501
01:09:42,800 --> 01:09:52,333
envisioned and designed for the low volume producers on campus. Those
that are doing, you know, an hour or four a week as opposed to 8 and

502
01:09:52,333 --> 01:10:00,999
10 hours a day. The system could handle something like that but we were
going after low hanging fruit and our observation was that a lot of

503
01:10:01,000 --> 01:10:12,466
departments were doing, you know, as I said one or two videos a week, you
know, or, you know, a dozen or so a quarter and that was about it.

504
01:10:12,466 --> 01:10:22,399
So, you know, those were the ones that we were going after there.
Angella mentions that she is going to be talking about some copyright
issues

505
01:10:22,400 --> 01:10:32,600
later in the presentation. So I will defer to her for more information
on that. Somebody asks are there transcription editors that stand for

506
01:10:32,600 --> 01:10:43,000
recommend CEUs for editing or creation of transcripts? I don’t really
want to say one or the other. What I can tell you is that I am working
with

507
01:10:43,000 --> 01:11:01,233
four right now. They are Video Jump, Cogi, Docsoft themselves and
another group called Project readOn or Blue Rhino I believe. All four

508
01:11:01,233 --> 01:11:11,399
companies have given us good service and we have been very satisfied with
them. The system is set up such that you can choose a transcription

509
01:11:11,400 --> 01:11:22,166
vender and is agnostic like that. The organizations we are dealing with
right now two of them are located in the Bay area and all four are

510
01:11:22,166 --> 01:11:31,232
located in California. So there was sort of that geographic proximity
that was of interest to us as well. Again this is all being handled on
the web.

511
01:11:31,233 --> 01:11:40,766
So they could be located in Anchorage, Alaska, for all I care. But it was
useful for us to have local venders simply because ...

512
01:11:40,766 --> 01:11:50,166
you know, we were spit balling this and putting it together from the get-
go. And so it was easier to have somebody local that we could talk to on
the

513
01:11:50,166 --> 01:11:59,832
phone. Somebody asked about moving it to Python or Jango.   Rob, yes, you
can e-mail separately on that. I would be more than happy

514
01:11:59,833 --> 01:12:04,733
I'm really agnostic. I don't care how it's done. What I would really like
to do

515
01:12:04,733 --> 01:12:12,633
is see the system move in to or at least some of the business logic that
we created move in to an open source delivery mechanism. Because I

516
01:12:12,633 --> 01:12:24,299
think, I think there would be value there and I am really big about
sharing that. So .. yeah. Somebody posted the question have you tried

517
01:12:24,300 --> 01:12:34,300
automatic sync. Yes, as a matter I am very familiar with automatic sync.
I am good friends, I know Kevin Erler very well and I talked to Kevin

518
01:12:34,300 --> 01:12:36,400
early on.

519
01:12:36,400 --> 01:12:47,300
And so going back to some of those commercial productions of, you know,
movies that disability resource center might have to do, I would
certainly

520
01:12:47,300 --> 01:12:56,833
say automatic sync would be a commercial service provider that would be
able to facilitate that kind of captioning.

521
01:12:56,833 --> 01:13:08,466
Again our system was envisioned to be used for web delivery primarily.
So I see it is roughly a quarter to the hour and I think we were supposed

522
01:13:08,466 --> 01:13:22,199
to be taking a brief break at this point in time. So unless there is no
other questions I am going to hand this back to Soji or Howard, I am not
sure

523
01:13:22,200 --> 01:13:31,600
who, but I am going to hand it back to the facilitator of the webinar and
I will be around throughout the process and able to answer

524
01:13:31,600 --> 01:13:42,800
questions later on in the seminar. So thank you to everyone.    Again my
e-mail address, I will very quickly get the slide up here, is
525
01:13:42,800 --> 01:14:26,166
jfoliot@stanford.edu and more than happy to answer questions via e-mail
as well. So don’t be shy and thanks once again.

526
01:14:26,166 --> 01:14:34,799
Soji: Thank you, John. And just want to make sure everybody -- I just
want to tell you we will be taking a short break for a couple of minutes.
So

527
01:14:34,800 --> 01:14:42,000
please stay online for the next presentation. Thank you.       (Break)
(We will be restarting the seminar in 20 minutes)    (We will be

528
01:14:42,000 --> 01:14:53,833
restarting at 12:05 p.m. MST).