ROUGH EDITED COPY

RIT/NTID

An International Symposium

Technology and Deaf Education

"Meaning-for-Meaning Speech-to-Text Services:

A Better Understanding"

Presenters:

Cindy Camp and Pam Francis

Session M10E

June 23, 2008

CAPTIONING PROVIDED BY:

ALTERNATIVE COMMUNICATION SERVICES, LLC

P.O. BOX 278

LOMBARD, IL 60148

* * * * *

This is being provided in a rough-draft format.

Communication Access Realtime Translation (CART) is provided

in order to facilitate communication accessibility and may

not be a totally verbatim record of the proceedings.

* * * * *

>>PAM FRANCIS: Good morning. Thank you for coming this

morning. Do any members of our audience need an

interpreter? Yes? So all right. Well, then, thank you for

being here, and I would ask, if you would like to introduce

yourselves.

>> My name is Miriam.

>> My name is Erin Jacobs.

>>PAM FRANCIS: Thank you. I am Pam Francis, and I am one

of your presenters today and this is Cindy Camp and we are

here to talk about meaning-for-meaning, speech-to-text

services and to try to shed a little bit of light and share

some information with you. We do welcome questions during

the presentation, and we're a little short on time already,

so we'll -- there may be some sections we speed through a

little bit.

Briefly, we're going to talk just about a definition, a

general definition, some of the benefits of

meaning-for-meaning, give some examples, and make an

association with some -- a process that might be familiar to

some, and some strategies that service providers can use to

improve their skills. And we'll talk briefly about analysis

and evaluation.

So if I was to ask the members of the audience what would be

your description of meaning-for-meaning, do you have

something in mind? Perhaps a summary of the spoken English

content. But that's not really what we feel is an accurate

description. A translation or paraphrase of the spoken

English content. We went a little bit further. And I'll

give you a chance to read that.

So we'd like to offer this as a definition that you can take

away with you today of what meaning-for-meaning

speech-to-text services are. And where did we come up with

that, or where -- what do we think is important as part of

that definition?

Well, we use the word translation, and we also use the word

paraphrase. And again, I'd like to give you a minute to

read the slide.

So as we move through our presentation today, you'll have --

you'll be able to associate what we're saying with the

definition that we have provided.

What are some of the benefits of a meaning-for-meaning

representation? Well, it is a model of written English. So

what does that mean? Well, if a speaker -- and I'll use

myself as a good example -- a speaker doesn't necessarily

speak in correct English, and a meaning-for-meaning service

provider represents the information in clear grammar, clear

format, clear English of what written -- a model of written

English rather than spoken English. Succinct delivery of

spoken information. Well, that might mean if something is

repeated, then the information might be bolded or it might

be underlined rather than put down twice.

A manageable transcript. For an hour class, rather than

receiving 25 to 30 pages of notes, they might review ten or

12.

Includes pragmatics of spoken information. And we'll talk

about that a little bit more. But the words don't always

represent what a speaker's meaning is.

And then it's visually accessible, meaning the information

can be bolded, it can be underlined, it can be emphasized.

There are many different ways to manipulate the information.

So, let's talk a little bit more about words versus intent.

As I said, linguistic meaning does not always reflect a

speaker's meaning. So let's look at your examples.

If I were to type, "Can you see Mary? " That might not give

a clear indication of what I'm trying to say.

So I may choose to use a different word. It depends how the

speaker is saying that information, what meaning they want

to convey. Because can you see Mary is very different than

"Can you see Mary? " Or "Can you see Mary? Can you imagine

what she's doing? So one might be a question as to whether

you can actually see her, and one might be a question as to

can you imagine what she was doing?

Again, another example, sincerity versus sarcasm. The door

is over there, I might be directing someone, whereas I might

be telling someone where the door is because I'd like them

to leave. How that information is conveyed is important.

And that's where the pragmatics come in. Words do not

always equal intent.

>>CINDY CAMP: How many of you, from being here, you've seen

what CART or verbatim transcripts look like. How many of

you have experience with meaning-for-meaning transcripts?

Okay. A few of you.

What happens is if someone has seen a service provider who

is not very skilled, then they come away with the impression

that verbatim is always best, and meaning-for-meaning is a

substandard service. And I'd like to challenge you today to

think outside the box and realize that these are two

different services. And it's very similar to thinking about

an ASL interpreter as opposed to a signed English

interpreter. They both have their place. It depends on the

client who's using the service, and the presenter, and the

situation.

I'm going to give you just a minute to read this transcript.

It's a verbatim transcript of a person talking about

transition services.

This is a very good example for our hearing audience because

when you read this and there is no voice saying the words,

much of the meaning is lost.

If you were hearing the information, it would make a bit

more sense. Still not perfect, but better.

And this is a very good example of when someone says, oh,

verbatim, that's always what we need, if a student's sitting

in a classroom and reading this, how much information do you

think they're actually getting?

It doesn't make a lot of sense. This is exactly how the

speaker talks, but as we've discussed, speakers don't always

use proper grammar or complete sentences.

So, my point is, just like ASL is not bad English,

meaning-for-meaning is not bad verbatim.

In fact, the mental processes that a meaning-for-meaning

service provider uses is very similar to those that an

interpreter into sign language uses.

If someone is using a verbatim method to transcribe, there's

actually very little mental processing that occurs in the

use of the language.

You may process homophones such as T-H-E-R-E, T-H-E-I-R,

words that sound the same, you would need to choose which is

the appropriate word, but other than that, it's basically,

the sound comes in, the words go back out on the computer.

With meaning-for-meaning systems, such as C-Print and

TypeWell, we're processing more at the phrasal and sentence

level. So instead of hearing a word and typing that word,

we listen for a concept, a complete idea, and then process

that whole sentence, that whole concept at once.

Looking at some interpreting models can really help us at

this point understand how meaning-for-meaning is processed.

These are three quite famous models, if you know anything

about interpreting. And we're going to briefly look at each

one of them and see how they apply to a meaning-for-meaning

service provider.

The Colonomos Model, I've simplified it for you. But

basically, she says that the interpreter listens to the

source language, mentally processes it, and decides how to

output that into the target language. And she broke it down

into steps.

This is exactly the same thing that meaning-for-meaning

service provider would do, listen for an entire concept,

mentally process it, think about what the best

representation would be. You can't always give out a word

for word translation because our students may not understand

that. The meaning may not be clear.

Pam gave us some examples of words that -- of sentences that

look exactly the same, but have very different meanings. So

we have to think about that.

Also, meaning-for-meaning service providers include the

environmental cues, tone of voice and inflection and how

those influence the output.

I have a really good example. There was a student who was

taking a class, and the instructor had a reputation for

being very difficult. So right before the midterm test, all

of the students in the class were scared to death. This

teacher's really hard, our midterm's coming up, we're all

going to fail.

The teacher REALIZING that they're all stressed out says,

oh, yeah, you all need to be worried. Everybody in here's

going to fail.

All of the hearing students went, whew, it's okay.

The deaf student in the class who had a CART service

provider saw those words, left the class and dropped because

they didn't get the tone. They didn't understand that it

was a joke to make everybody feel better.

The second model I want to talk about is the Gish Model, and

this one is actually one I find most helpful because what

happens with most service providers doesn't matter what

language they're going with, they get stuck down here on the

detail level.

1492, Columbus! What about it? If you don't have those

connectors that Columbus discovered America in 1492, it

really doesn't matter that you got the year and you've got

the name.

So it's important to realize that we need the goal of the

speaker, and we always need to keep that in the back of our

heads.

What is the goal that this speaker wants to convey today?

Then we go down, there's going to be a theme. You'll have

several objectives. Then you get down into the units and

the details.

If you always keep going back up to that main goal, you're

pretty safe. I joke about if you're interpreting in church

or captioning in church and you lose the content, what's the

overall goal in church? God is good. Everyone should go to

church. You go back up to that main goal.

Another example from a classroom, a captionist went in, and

the teacher said, we're going to show a video today. Does

everybody love trying to interpret or caption for videos?

No.

It's impossible.

So the captionist went to the instructor and said, what is

the goal? What do you want them to learn from this video?

I'm not going to be able to keep up with the dialogue and

get it all, I'm going to have to summarize, so if you let me

know what you want them to learn, then I can focus on that.

The teacher said, the sound effects.

The captionist said, the sound effects?

Yes. It was a war movie, and this was a class where then

they were going to talk about how those sound effects

enhanced the plot and so on.

So the captionist dropped all the dialogue and focused on

the sound effects.

At the end, when the class was discussing the movie, the

student was able to fully participate, whereas if the

captionist had not asked what the goal was, she would have

focused on the dialogue, and then the follow-up

conversation, the student would have been totally lost.

So even had she been able to get verbatim of the dialogue,

that still wouldn't have helped. So that's why I find this

model very helpful.

This is the third model created by Vance Cokely, and I'm not

going to go into it at all because it's a very detailed

model of the mental processes involved in interpreting. But

I want to tell you that this is very similar to what a

speech-to-text person experiences, so this is your brain on

speech-to-text, and meaning-for-meaning. It's not an easy

thing.

The second part of the Cokely Model talks about how the

service provider influences the message sometimes

intentionally and sometimes not. And we can talk about how

things can be misrepresented. And again, this is the same

in meaning-for-meaning.

There are things that are omitted, there are things that

could be added that the speaker did not intend.

Sometimes substitutions with words that are not appropriate,

sometimes environmental intrusions into the service

provider's mind.

There's a cell phone ringing, there's a student who keeps

tapping their foot and the service provider misses

information because of that. And then sometimes there are

just those weird things that show up in the transcript that

we didn't intend.

In meaning-for-meaning, that can be a word that did not

expand correctly in the software. Just things like that.

When you're looking at a verbatim transcript you're going to

get basically the same thing. But I wanted to show you,

with meaning-for-meaning you're going to get variation, but

that's not necessarily bad.

Some service providers may provide more detail than others,

depending on skill level, depending on the content.

But I have an example of the verbatim of a lecture, and then

two different meaning-for-meaning transcripts. So you can

see what you might get. And I'll give you a minute to read

those.

Okay. After seeing an example of a meaning-for-meaning

transcript, what's your impression?

Do you think it's worse than verbatim? Do you think it's

appropriate in some situations? Everybody's still asleep

this morning. But I see some nodding.

Word-for-word transcript. Every single word. Everything

that's said, it's that.

Okay. Hopefully now you have an understanding of the

difference between verbatim and meaning-for-meaning, and

hopefully you can see that each has its place. If you have

an instructor who does not speak in complete sentences, and

rambles, then perhaps the student would prefer

meaning-for-meaning.

We had a situation where the instructor very clearly

outlined everything and was very organized, and the student

normally had a meaning-for-meaning service provider. We had

a substitute person come in who used CART, and the student

said they really preferred meaning-for-meaning in that

situation because of the formatting that could be shown.

In meaning-for-meaning systems, you can use bulleted lists,

numbered lists, bold, italics. It's very visual. And when

the instructor isn't -- speaks in an organized fashion and

uses those formatting tips, then the student was able to

follow along.

Whereas the CART provider used a paragraph format and didn't

have all of the formatting available to them. So I

challenge you to look at the situation and see what is going

to accomplish the goal of getting the information across.

Yes? We have a question?

>> I'm just wondering, with those two examples, do you have

any comment about the quality of one over the other?

>>CINDY CAMP: I think they are both very good examples. I

did not -- we have some examples later of where information

was missing and where things were lost. These are just

meant to show that there can be differences.

This is just a list of some skills that someone who was

providing a meaning-for-meaning service would need to have.

Very strong short term memory because it's not going in and

out. It's going in, process, then being put out. At the

same time you're listening for the next chunk of

information.

You need an expansive vocabulary. If there's a word that

could be confused with another word, what do you have in

your vocabulary to replace that?

Very good comprehension of the English language,

understanding idioms and knowing that those are probably not

going to translate well for someone who has English as a

second language.

A good knowledge of deaf culture and syntactical references.

Yes?

>> Is there a case where you may have the service provider

provide the idiom and the true meaning so they still get

exposed to that group of vocabulary, or in context of what

the speaker was saying?

>>CINDY CAMP: It's going to depend on the student and the

service provider. The majority of the time, yes, they will

give both. However, I've had situations where it disrupted

class greatly because the idiom was given and as soon as the

student read it, they burst out laughing because they had

never been exposed to it, and then the instructor thought

they were laughing at him, and so you have to be very

careful about how you do that.

You need a broad knowledge base. If you don't understand

the content that's being expressed, there's no way you can

translate it into another language. And trust me, written

English is a different language than spoken English.

If you've ever sat in a classroom and heard the instructor

lecture, then they read from a book, the syntax, the rate of

speech, the structure, it's all very different. So you have

to understand the concept to be able to express it

appropriately. You have to be very strong at summarization.

And summarization doesn't mean contents left out. It means

it's presented in a clear and concise fashion.

And then obviously you need to be a fast typer.

I'm going to turn it back over to Pam now and she's going to

talk a little bit about how you could evaluate a service

provider that you're seeing or at your institution, and then

we'll have a few minutes at the end for questions,

hopefully.

>>PAM FRANCIS: The evaluation of meaning-for-meaning could

take -- the discussion could take a few hours. So we won't

dive very deeply into it, but we do want to briefly address

it. There are existing systems out there that evaluate

non-verbatim representations of information.

For example, foreign language certifications, that might --

that's one example.

And there are systems that are working, they're in court,

but to get there is a very, very complicated process and

time intensive.

So what do those certifications systems, those evaluation

systems have in common? A specified unit of analysis. For

example, idea units, certain chunk of information broken

down. Objective scoring criteria, very, very important

because as Cindy said, you can put five people in a room and

they can transcribe the information, and what you get at the

end looks different from all five people. Okay. So you

have to have objective scoring criteria.

Consensus scoring. For these systems you need something

that more than one person evaluates the output. That is

very important, and it's very important to really reach an

objective evaluation. And they want the final outcome, the

final outcome needs to be valid and reliable. This is a

quantitative system. It's not -- it's much more objective.

It's much more precise.

But to get there takes quite a bit. Five minutes? Okay.

To get there takes quite a bit. And this is just really a

representation of what it might take to develop one system.

Preparation of the source material. Getting the lectures or

getting a piece of spoken information that has a certain

level of content, and then recording it, so on and so forth.

Identifying the units of analysis. So breaking it down into

idea units. That's a huge task. If you're talking about a

one-hour lecture, you can end up with 300 to 400 pieces of

information or idea units.

Training judges, training people who will take that

information and evaluate that transcript. And that's not

just one judge. There may be three. There may be five. It

depends on the system.

Performing the analysis. Actually having the judges go

through idea unit by idea unit and deciding whether or not

it's appropriate. And then summarizing and reporting. So,

for one, okay, we're talking 38 to 42 hours. That's a lot

of time that most -- unless you have some system set up to

evaluate, for example, you have a company that's doing the

evaluation, you have some system that does evaluation,

that's a lot of time. And most places, especially campuses,

do not have that kind of time to devote to that evaluation.

There is an effort to try to develop some type of national

certification that is appropriate, but that -- to develop

the system, that is legally defensible, that is objective,

it will cost well over a million dollars. So what do you do

in the meantime? And I'm going to be very quick about this.

You have to look at the real time transcript and decide

whether it has the qualities that are needed.

Again, you have five different transcripts. You have to

look at them each individually. You have to decide whether

things are clearly worded, whether the information is

complete and accurate, and whether it's easy and quick to

read. And that doesn't mean that it's at a lower

educational level. That means that the reader can look at

it, get the information and move on, that it's not

convoluted in any way, short, quick sentences representing

the information.

I apologize that I'm running along.

So, how do you do that? You get the unedited transcript.

Okay? And you review it.

It would be great to be in the classroom or in the situation

where you're hearing the lecture and then you can go back

and review it. You don't always have that option. Some

things you can do is get notes from a notetaker. You can

review the transcript with a consent expert or the

instructor. There are a few things you can do.

Okay. Check the transcript out. This is just a few

examples. One thing that drives me crazy is when I see a

transcript of just solid text. There's no visual breaks.

There's no -- the information is not clearly represented

where a thought starts and stops. So really look at it and

make sure that the content is good and the representation is

good.

Once that transcript is reviewed, there are things that can

be gleaned from that, and information -- assistance can be

offered to the service provider.

For example, depending on how they're representing the

information, perhaps giving them suggestions, and we passed

out that strategies for service providers sheet. On how

they might include their wording or their vocabulary, how

they might better capture the information.

So really, what is gleaned from that review is very

important, and that information can be shared with the

service provider, and their skills can be improved.

This is a very basic one-hour, two-hour process. It's -- at

least for the review.

It's not meant to be a legal process or even a process that

can be used to hire or retain a person unless it's done at a

very -- much deeper level. But the point is, is the

interview -- or the transcript can be reviewed and

information can be provided to the service providers so that

their skills can improve.

As far as the certification or evaluation system, that is

widely available. That takes time. It's not easy to take a

meaning-for-meaning transcript and objectively say, you have

all the information, you have 60% of the information, you

have 30%, you have 90%. It's very difficult to do. And

it's time intensive. But there are things in the interim

that can be done. So what did we talk about today?

We gave a brief definition. We talked about some benefits.

Cindy talked about interpreting and the interpreting models

and things that might be familiar, and some skills that

service providers, meaning-for-meaning service providers

need. And a very brief discussion about analysis.

Are there any questions at all?

I think what's -- oh, go ahead.

>> I'm interested to see the meaning-for-meaning. In the

U.K., it is in the training course for notetakers, but I

found that it takes a lot longer for people to learn how to

do it, especially if they have fast typing speeds they tend

not to be good at summarizing. Is this part of your initial

training for a notetaker or is it an add-on?

>>PAM FRANCIS: Well, one of the things that is an issue

here is that the meaning-for-meaning service providers are

very concerned, and it's not notetaking. If they're

accurate and they -- and they get all of the information,

it's not really notetaking. And depending on the skills of

the student -- bless you -- depending on the skills of the

student, if the student needs notetaking and the information

that a meaning-for-meaning service provider gets is too

much, if they're overwhelmed, then they have to cut back.

So what we do, at least in the training that we -- that we

have done is we provide some skill exercises, we provide

information about listening, so on and so forth.

I think one of the very difficult parts is that

meaning-for-meaning service providers, one of the reasons

that they were established is because they wanted, and I

apologize for talking very quickly -- they wanted a training

that was quick, and the people could get it into the

classroom.

The problem with that is that you cannot develop the

processing skills that are necessary in three weeks or four

weeks or five weeks, in six months.

Okay. So that is the problem or a main issue is developing

the processing skills to be able to take the information and

put it back out. Many meaning-for-meaning service providers

don't come in with that. It takes time. There are many

people who are cross-trained. The best meaning-for-meaning

service providers are people who have those skills, like

interpreters.

>> Thank you, Pam and Cindy! Thank you very much for your

presentation! I'm sure there are many of you with

questions, and you can approach Pam and Cindy during the

rest of the conference.

We are very interested in your feedback. We are encouraging

you to fill out evaluations online, and the learning center

is available for that.

I have some hard copy evaluation forms if that is your

preference. Each session is named for the day and the time.

This session is M10, Monday at 10:00, and that's the most

important thing when you're providing feedback. If you

would like to provide written feedback, please take one of

these forms and complete that. Thank you so much for

coming!

>>CINDY CAMP: I have some brochures on the speech-to-text

services network, which is an organization for both

meaning-for-meaning and verbatim. I don't have enough for

everyone, but you're welcome to take them and go online and

learn more.