Multimedia Captioning by Automatic Sync Technologies: Transcript

ROUGH EDITED COPY
ALL RIGHTS RESERVED
UNIVERSITY OF ILLINOIS
CIRCLE CAMPUS
DIGITAL ACCESSIBILITY EXPO
NOON SESSION C
MULTIMEDIA CAPTIONING
BY AUTOMATIC SYNC TECHNOLOGIES
1:00 p.m. - 1:30 p.m.
APRIL 9, 2009


CAPTIONING PROVIDED BY:
DILLON REPORTING SERVICES
100 NORTH LA SALLE STREET, SUITE 1500
CHICAGO, IL 60602

*  *  *  *  *

This is being provided in a rough‑draft format.  Communication Access Realtime Translation (CART) is provided in order to facility communication accessibility and may not be a totally verbatim record of the proceedings.

*  *  *  *  *

>> SPEAKER:  Pat Brogan from the Automatic Sync Technologies is here.  She's had over 25 years of experience in the field of assistive technology and she's going to talk about some of their programs and what they provide for captioning.  She'll go into it.  I'll let her go into it and talk about it. 
>> PAT BROGAN:  Thank you. 
How many of you do any captioning now?  How many of you understand, separate from section 508, what the value, why you might do captioning?  Anybody have ideas? 
>> AUDIENCE MEMBER:  For any student, it's a way for them to be exposed to the information and multi sense.
>> PAT BROGAN:  Okay.  Any other thoughts about why you would caption? 
>> AUDIENCE MEMBER:  If they couldn't play the sound.
>> PAT BROGAN:  If you're in the gym and you're studying and you ‑‑ you got to have that turned off or if you're on an airplane.
>> AUDIENCE MEMBER:  If a person has a heavy dialect, sometimes it's hard to understand them.
>> PAT BROGAN:  That's true.  If it's a faculty member.
Other ideas? 
>> AUDIENCE MEMBER:  Background noise. 
>> PAT BROGAN:  Background noise.  That's right.  Yes. 
>> AUDIENCE MEMBER:  (Inaudible) someone mentioned earlier, but if it's done earlier it can be done to facilitate search.
>> PAT BROGAN:  Search is a really good answer and it's discoverability and search. 
C net captioned all the video on their website and overnight their Google hits went up 30 percent because they were able to get down to the actual content and bring visit stores to the content that was ‑‑ visitors to the content that was relevant to them.
I'm going to talk about captioning.  What I have here is a recorded lecture that was captioned, and before I start, you know, talking about the whys and hows of our company, and the captioning process, I wanted you to think back about the topic of universal design that people have talked about.  And if the ultimate goal is understanding how students learn, then one of the technologies today that's being used is mobile learning, and tools like this.  This is a workshop that I participated in at Gallaudet university.  And what this is is it's a capture of five speakers who talked.  We had the former chairman of the board of Gallaudet University, we had Gallaudet University, George Washington University.  I spoke, a representative from echo 360.  And this is the output, this is being streamed live from a server in Virginia of the recording that was done and it's similar to recorded classrooms that many campuses are doing, including the University of Illinois on the various campuses.  Not all the classrooms.  But the idea behind this is you can navigate ahead.  This is the section on captioning.  It has audio.  In this case Earl is doing the presentation with a CART service similar to what we have here.  The captioning is done not on the audio necessarily, but sometimes you can ‑‑ well n this case, we did ‑‑ in this case we did caption the audio because we recorded that but we also had the signing that was done.  So for students who are learning, they can make this full screen.  Make this full screen.  They can watch it and skip ahead.  This is being delivered as a flash movie and as the gentleman from adobe said the fact that you can ‑‑ if you have captions associated with flash you can search throughout that and come back to the point of a lecture. 
So the reason I'm showing you this is because students today are used to learning in a way that meets their needs and it may be to watch the professor, it may be to read the slides, it may be to listen to it, watching the signs as they're made if they don't have hearing capability.
So our company was founded by a Department of Ed grant in 2003.  The founder of our company's background is in speech recognition technology.  And I say that, but our company does not use speech recognition technology.  Part of his grant funding was to look at can you take the technology of speech and make it acceptable to doing transcriptions and captions with a quality that meets both 508 and 504 standards and intelligibility in academic settings and the answer is we're not there yet.  So part of it is how can you automate the process assuming there are going to be mandates that say you should be captioning all of your audio and video that gets delivered to the web, and separate from the mandate of why you should do it, it really makes good sense to it do it for a lot of different learning styles.
We talked about why caption and there's a lot of challenges around complying with 508 and 504 of the Americans with Disability Act because the concept is a good concept that most people believe has a lot of sense but the fact is without having automated tools, you still have to go through a process, and it does cost money.  And so that is one of the challenges. 
Somebody did mention ESL.  And faculty who we looked at, the reasons that student watched ‑‑ students watch recorded lectures, often they've missed a class, they want to review a section, but English as a second language professors is a huge issue.  They may not recognize ‑‑ they may not be able to read their handwriting or they may not recognize their voice.  And a lot of times, they may not even be able to interact in the class with questions because they don't understand what the professor is asking.  And with students today, there's research that says students read on average a total of aggregate of seven books a year, so language skills, written language skills are getting to be a problem and in academia, to understand key terms to be able to test on them appropriately, the visual recognition is really important and I always go back and cite (inaudible) which says if you read it and see it and hear it and apply it to multiple challenges of the brain the likelihood that you will retain it goes up significantly.
And now content is being rendered more mobile.  Students are listening to lectures on iPods, iPhones and taking advantage of devices they have in mobility. 
We talked about searchability and so as lectures become more reusable the ability to and a half navigate exactly to the part where the professor said we're going over non‑linear equations today and here's the example, if you didn't get it in the classroom, here's an opportunity to do that.
I'll give you an example of Arizona state university has 2,000 students with disabilities and they have been sending note takers to class for all of these students for years.  The concept now to record the lectures, to caption the lectures, and let tutors actually work with the students at the point of need rather than automatically sending note takers to the class has ‑‑ is a part of their future plan, and the students love the fact that they can watch them over and over and if they need help they can go ask a tutor and say take me to that section right here or they can go back to the professor. 
So these are the why captions.  What we found in our research was that we were going to look at the techniques and cost structures because captioning has been typically a very expensive and it still is expensive in the overall scheme of campus budgets and we concluded that significant improvements were able to be achieved in the process of captioning, but that you really needed live bodies to develop an accuracy rate that was acceptable.  And the surprising results were that we looked at error rates, and comprehension.  This passage is correct.  A hundred percent, zero arrow rates. 
This next passage has a ten percent error rate.  And if you try to read this, what happens?  And we're talking ten percent.  Speech recognition tools frequently have 70, 80 percent rates. 
When you get to 20 percent error rate, you can totally lose the context, and so what happens to the brain?  The brain turns off.  So the research was designed to really look at this and say what is an acceptable error rate, and it varies by individual and it varies by discipline and topic, that you could use to then automate the process, and be able to develop, deliver back quality output.
What we found is that there's a dramatic drop in intelligibility and error rates, and I'm not going to go through all of these.
So what did we look at?  We looked at what are the options that we can use to train captioners.  We can use inexpensive labor, students and a lot of campuses do use students to do transcriptions.  We can use speech to text solutions.  And those are the ones, and there are product out there in the marketplace, just like the gentleman from adobe was talking about, and they can be trained and as you train them, the accuracy does go up.  But the question is where do you get to, and there is a point of diminishing returns where it was far easier and less expensive to transcribe it right from the beginning than to go back and correct it. 
So these are typical error rates.  And I've heard this woman here is the best on campus and has no error rates and so people were saying they want her in their classroom. 
(Laughter.)
I don't know where this came from. 
Student transcribers, that may vary depending on the consistency of the pool of labor that you have, their knowledge of disciplines, and there are things you can do by equipping people for specific types of transcriptions with medical terms, glossaries, et cetera for people to do this professionally. 
Now, speech recognition trained, I've seen one campus that I've worked with get in the mid 90 on an IT course with a native English speaker who has gone back and done training with automatic speech recognition tools and then untrained, it's 20 to 40 percent error rates. 
So our conclusions were that a woman like Jo sitting up here are an essential part of the transcription process and our company has actually automated the process of creating transcriptions and captions, but we use human transcribers, and court stenographers to actually do the work. 
So what we did is we focused on the work flow, and the work flow is that we will take an audio file or a video file or a DVD, if it's a DVD, obviously it gets physically delivered to us, but if you have an audio or a video, that you would like to have captioned or transcribed, you can log on to our website, upload the file.  If you have your own transcript.  So Jo is creating a transcript right now.  If that transcript were uploaded to our system, within ten minutes a captioned file will be returned.  So the captioned file can be used in a number of different ways.  If I were being recorded and a movie were being generated of that, the captioned file would sync the transcript with my audio and the video, and that's essentially what a captioned file, you will sync it to both audio and visuals. 
So what we do is we have a system where you can upload the materials, we will either do the transcript and/or the caption.  We will give you a choice of options, so you may have as many as 40 different output types depending on the purpose of your content.  So if you want flash or winds owes media or ‑‑ windows media or real. 
If you're using electric capture solution like I showed you in the Gallaudet example, there's an automatic published to our company built into the system.  So if you're scheduling a classroom recording or doing a ad hoc recording you can say automatically send the MP3 file to us.  We generate the transcript and the captioned file.  We redeposit that back into the server, it gets synchronized with the video, the classroom visuals and the audio, and then delivered in that window.  The associated transcript file can be appended to that.  So if you have for example I was at Arizona state fairly recently and I was talking to a student in the Cronkite school of journalism, he is blind and relies heavily on transcripts to use with jaws for his classes.  Even though he has the audio, having the transcript with the audio gives him more control over using the speedup tools and the system a little bit better. 
So these are some of the output formats that we support.  We also give you up to six months to redo them to edit or generate more output files at no additional charge.  So if you have a five minute video that you have that you want to publish to a website, by law that should be accessible and captioned
.  It costs $14 to do a five minute video.  We charge in five minute increments based on an hourly rate and the more you do the lower the price goes and the idea behind it is we will have production simplicity.  So it's web based.  You can upload the files, get them back automatically, and if you go to our website, you want to see how to caption flash or windows media, there's a lot of how to movies that show you how to do that or how to get captions into a DVD. 
And that's pretty much all I was going to talk about.  Does anybody have questions, comments?  You all ready to caption everything so you're compliant?  Do you understand what Missouri just did in terms of their 508 compliance?  They actually added treble damages in punitive dollars associated with any person who delivers content that not accessible to the web.  So Illinois is a good plan but I think Missouri is one that has punitive damages. 
Well, thank you. 
I'm next door if you have questions. 
(Applause.)
>> KEVIN PRICE:  Thank you very much. 
We're going to have the web master forum here in a few minutes.  Just we have time to stretch your legs and get everyone here, so it will be a couple of minutes and we'll have the web master forum started.
(End of session.)