Phew - thanks for the heads up on this, Peter. I wasn't on the
foundation list until a few hours ago, so I missed David's original e-mail.
* Perfect a free software eye tracker program like OpenGazer (needs a
*lot* of work to be usable & stable)
There's also MouseTrap, which is based more on the simpler problem of
head tracking. It was a successful "GNOME Outreach Program:
Accessibility" project, and is also something being continued as an
HFOSS project. So, it might already have enough coverage. It would be
nice to see if the OpenGazer and MouseTrap folks could collaborate on
code or algorithms via the nice open source community we have.
* Open voices - doing quality synthetic voices is a lot of work, major
research project & lots of time in a sound studio with specialised
actors. Funding one (or several) in various languages would be useful.
There are a few different ways to tackle speech synthesis. The one you
allude above is generally related to concatenative synthesis, and is
what you get with systems such as Festival, Cepstral, etc. The voice
data tends to be very large, sometimes on the order of 100's of Mb for
the more natural sounding voices. While the voices are very natural
sounding at normal speaking rates, they tend to become less intelligible
at the higher speaking rates blind users are accustomed to. In
addition, the performance of these voices (e.g., time to speak, time to
stop speaking, etc.) can be poor. Finally, getting good data not only
requires picking the right voice talent, but also requires many hours of
aligning the recorded data properly so as to get good units. It's very
expensive.
Another kind of speech synthesis is formant synthesis, and is what you
find in systems such as DECtalk, IBMTTS/ViaVoice, and eSpeak. These
voices are less natural sounding, but are more intelligible at higher
rates of speed. A good friend of mind, who also happens to blind, often
gives the analogy that these engines are the equivalent of fonts: they
obviously aren't made by humans, but they are a lot easier to read quickly.
As Peter mentions, there is AEGIS work to expand the language coverage
of eSpeak. So, I think a prime candidate for work might be to work with
the eSpeak author (Jonathan Duddington) to do something such as allow
pluggable vocal tract models and then work to decrease the 'harshness'
of the engine that users have complained about. This would improve the
acceptability or 'listenability' of eSpeak, while the AEGIS work would
improve the locale coverage. With these two things, we could
potentially get a very good open source synthesis solution that is
small, fast, and intelligible at high speaking rates for a lot of the
world's languages.
Internet + accessibility:
* Integrate an eBook library like Gutenberg Library or Bookshare into
the desktop - integrate well with Orca to make a book reader
The good folks at Benetech (creators of Bookshare) are using Mozilla
Foundation funding to create a DAISY (and eBook) reader as an extension
to Firefox. Called DAISYfox, it was demoed at the CSUN Conference on
Technology and Persons with Disabilities last month, and is coming along
nicely.
I helped get DAISYFox off the ground via Mozilla grant funding and was
really pleased to see the results at CSUN. We've worked closely with
them to make sure that it works with desktop screen readers (e.g., Orca
is at the top of my list ;-)) as well as being self-voicing if needed.
After a great meeting at CSUN, we're looking to get additional funding
from the Mozilla Foundation to carry the work forward, and future work
involves several interesting areas, such as managing your local DAISY
library, integration/search for Bookshare and other content providers,
etc. If that falls through, however, I strongly believe in DAISYFox and
think it is a great candidate for a grant.
I also want to second Alberto's suggestion of GtkWebKit accessibility
support. This effort could use more folks engaged in it...
I've set up a WebKit "hackfest" for this Thursday, which is more of an
opportunity for us GNOME a11y folks to teach the WebKit folks how to use
GNOME accessibility tools such as accerciser and orca to analyze the way
WebKit exposes things to assistive technologes via the AT-SPI. My goal
is to "teach them how to fish", and to be able to compare their
implementation to the Gecko implementation in Firefox. From there, we
can ask them to scope out the work to support AT-SPI and ARIA and then
see how much work they think is necessary to do the job.
For other ideas:
Eitan Isaacson will be mentoring an HFOSS intern project for a braille
transcription package. This project will exercise the liblouis package
a fair amount and should flush out many issues and answer many
questions. From there, we can take this knowledge and apply it to
something such as doing braille transcription directly in OpenOffice via
a to-be-written extension. Having the transcription software directly
in the content generation tool is the right spot, IMO, and is something
that is probably worthy of a grant.
The Bonobo/CORBA deprecation for GNOME 3.0 will require a rewrite or
replacement for gnome-speech. Luke Yelavich from Canonical is looking
for support from his management for doing the work. Without support
from his management, we're going to be stuck. This is an interesting
project with good value and cross desktop (e.g., GNOME and KDE) opportunity.
The current on screen keyboard solution (GOK) could use some updating,
and perhaps there's an opportunity to make it support a larger audience
(e.g., mobile devices with touch screens but no keyboards) as well as
better integrated with the desktop.
The deafness and learning disability communities also don't have great
coverage in GNOME. The VizAudio work that Bryen Yunashko is mentoring
for HFOSS will hopefully yield some good results for the deafness
community, but more work would be welcome I'm sure. We can do some
stuff for the learning disability in Orca via highlighting and trimming
of speech output, etc., but there's nobody lined up to do the work right
now.
Finally, please check out our already-prepared page of GNOME
accessibility desires, at http://live.gnome.org/Accessibility/GetInvolved
Indeed. This is a good spot for ideas and has more than I mentioned
above. The GNOME accessibility community tries to keep this up to date,
but it would also be good to drop a message to
[email protected] before starting something or if you
have questions. It's a friendly group. :-)
Will
GNOME a11y lead
_______________________________________________
foundation-list mailing list
[email protected]
http://mail.gnome.org/mailman/listinfo/foundation-list
