[Athen] Synthesized Voices and Accurate Pronunciation

[Karlen Communications]

Some of the synthesized voices or maybe it is a function of JAWS, let you
adjust the pronunciation dictionary. In JAWS the keyboard command to open
the dictionary is JawsKey + D if you are using Eloquence. You can modify the
default dictionary of pronunciations the dictionary for a specific
application. I do this all the time as it also lets me shorten phrases that
are repeated in subject lines to a single word - for example if the ATHEN
messages came in with [Access Technology in Higher Education Mailing List] I
can enter that in the first field and then enter ATHEN in the second field
so that messages would say ATHEN and then read the rest of the subject line.
Until the Eudora e-mail client was added to the default list, we had to
enter it manually. So you would put Eudora in the first edit field and U
dora in the pronunciation field.

I think you can back up the dictionaries or save them if you install a new
version of JAWS.

One of the problems I have with the Victor Reader is that there is no
mechanism to work with the pronunciation dictionary and something as simple
as wasted is pronounced was Ted. Many of the commonly used words have not
been added to pronunciation dictionaries for the natural phoneme voices. I
find this distracting when I'm reading because I spend more time trying to
figure out what the text is so my comprehension level is quite low and I get
frustrated. On the other hand I know a lot of people love the natural
phoneme voices because they do sound more natural.

Check to see if you can modify or add to the pronunciation dictionary for a
set of voices. I don't think you can with the Microsoft ones.

Terry is right about DNS and they do have a medical version of the software
so the pronunciations might be in the TTS tool that ships with it.

Cheers, Karen

-----Original Message-----
From: athen-bounces at athenpro.org [mailto:athen-bounces at athenpro.org] On
Behalf Of Terrill Thompson
Sent: September-17-09 7:33 PM
To: Access Technology Higher Education Network
Subject: Re: [Athen] Synthesized Voices and Accurate Pronunciation

Ellen, 

The speech synthesizer also plays a key role in pronunciation. For example,
I just tested a block of medical text using various synthesizers. Eloquence,
the default JAWS synthesizer, didn't handle the medical terms very well at
all, but when I switched the JAWS voice profile to SAPI 5 (Microsoft Anna),
pronunciation was much better (though not perfect). 

There are other SAPI synthesizers/voices available, most for a small cost;
and most Windows screen readers or text-to-speech software products support
switching to these if they're installed. Some of these (e.g., AT&T Natural
Voices, Neospeech) have nice-sounding voices but they won't do any better
than the free Microsoft voices at pronouncing medical terminology. If using
a Microsoft SAPI voice and techniques like those described by Dan won't meet
the student's needs, 
what they need is a voice that is specifically equipped with a medical
vocabulary. I'm not sure what's available in that arena, although you did
mention Dragon Naturally Speaking - that might be worth exploring further.
Even though the student doesn't need speech recognition I wonder if by
installing Dragon Medical you would be installing a medical SAPI voice that
could be used in other text-to-speech applications. Does anyone know?
Unfortunately Dragon Medical is pretty pricey ($1600 through Nuance) - it
would be a shame to pay that much if you're only going to use one component,
not the application itself. 

If anyone wants to conduct their own tests, here's the medical text I was
working with: 

Methacholine Chloride is a quaternary ammonium parasympathomimetic agent
with the muscarinic actions of acetylcholine. It is hydrolyzed by
acetylcholinesterase at a considerably slower rate than acetylcholine and is
more resistant to hydrolysis by nonspecific cholinesterases so that its
actions are more prolonged. It is used as a parasympathomimetic
bronchoconstrictor agent and as a diagnostic aid for bronchial asthma. (From
Martindale, The Extra Pharmacopoeia, 30th ed, p1116)

Regards, 
Terry

Terrill Thompson
Technology Accessibility Specialist
DO-IT, Accessible Technology
UW Technology Services
University of Washington
tft at u.washington.edu


-----Original Message-----
From: athen-bounces at athenpro.org [mailto:athen-bounces at athenpro.org] On
Behalf Of Burke, Dan (DSS)
Sent: Thursday, September 17, 2009 2:23 PM
To: Access Technology Higher Education Network
Subject: Re: [Athen] Synthesized Voices and Accurate Pronunciation

The main determinant of accuracy in reading is not what is doing the
reading, but the quality of the electronic text that is being read.

For example, , JAWS out of the box pronounces the name of my town as
Miss-owl-a, when it is commonly pronounced as Mizzoola.

So the issue is for the student to learn a couple of tricks -- stop
(frequently, if necessary) and spell out words he or she is uncertain
of.  If they are pronounced strangely, virtually every tts package
offers some way to correct pronunciation.  Which is what I have done
with JAWS and the name of my town.

So, the quality of the e-text -- is it a publisher file?  If not, has it
been properly spell-checked?  

And then add to that the student taking responsibility (with our
guidance) and learning how to manipulate the screen reader to make him
or herself successful.

Dan


Daniel J. Burke
Assistant Director/Coordinator
Disability Services for Students
Emma B. Lommasson 154
The University of Montana
Missoula, MT 59812

www.umt.edu/dss/

406.243.2243 voice/text
406.243.4424 direct line
406.243.5330 fax


-----Original Message-----
From: athen-bounces at athenpro.org [mailto:athen-bounces at athenpro.org] On
Behalf Of CUTLER_ELLEN
Sent: Thursday, September 17, 2009 2:37 PM
To: Access Technology Higher Education Network
Subject: [Athen] Synthesized Voices and Accurate Pronunciation

Hello,

We have a student who needs to listen to respiratory therapy text books.
Towards this end, we need to know what determines how accurately text
will be read. Is it the synthesized voice? Is it the text-to-speech
software? What text-to-speech environment is likely to read medical
texts accurately? I was wondering if the TTS feature of the Medical
version of NaturallySpeaking would do well; however, the student does
not need voice recognition....

Thanks, Ellen


Ellen Cutler
Santa Monica College
Disabled Student Services, High Tech Training Center
1900 Pico Blvd.
Santa Monica, CA 90405
310.434-4496
cutler_ellen at smc.edu




_______________________________________________
Athen mailing list
Athen at athenpro.org
http://athenpro.org/mailman/listinfo/athen_athenpro.org

_______________________________________________
Athen mailing list
Athen at athenpro.org
http://athenpro.org/mailman/listinfo/athen_athenpro.org

_______________________________________________
Athen mailing list
Athen at athenpro.org
http://athenpro.org/mailman/listinfo/athen_athenpro.org