one step closer to the possibility of linux on the bn/vn

Tue Apr 15 09:20:43 MST 2003

Hi Sabahattin and all.

You raise some very good points. I must admit that I am bias towards speakup, so it's good to see someone express alternate views on the subject (smile).

While it is true that the bn is not supported under speakup, I believe that PDI is the one to blame here, since they don't want to release there specs without nondisclosure agreements.

Also, as for modularizing speakup, I believe that Kirk, the main developer of speakup, stated on the speakup list a few days back that the speakup 2.00 tree which will be in cvs shortly, will have the ability to build speakup as modules.

Also, as for IBM's via voice TTS, it is dead under unix, and the development has stopped as far as I've heard, so that's one less option for a speech engine.

Greg

P.S., I'll hunt up the URLs, and post them in another message.

On Tue, Apr 15, 2003 at 04:36:24PM +0100, Sabahattin Gucukoglu wrote:
> 	On 11 Apr 2003 at 17:42, Gregory Nowak pronounced thusly:
> 
> > Hello everyone.
> 
> Hi Greg and others,
> 
> > This list has been dormant for a very long time, so I thought I'd post on a
> > somewhat significant development.
> 
> I must be partially held to blame for this. :-)  For I have been 
> particularly busy of late and it has normally compelled me to "Leave the 
> email till later."  I hope I can make up by means of this and future 
> hopeful discussions.
> 
> > One of the problems with porting the linux kernel to the bn/vn was the
> > question of speech. Currently, under keysoft, the keynote software speech
> > is used through the sound chip.
> 
> Giving this some fairly analytical consideration, it turns out to be a 
> very big problem indeed.  If the Linux (or indeed any other Unix flavored) 
> ported BN/VN is to use the original Berkeley speech tech synthesiser (now 
> L&H Speech Products) used in the BN, a port of it from Windows 16-bit or 
> Windows CE code is first necessary.  The best chance we had in this field 
> was when the hardware forms of these synthesisers in chips were available, 
> as used in the original Keynote products.  Right now, though, it will be 
> an uphill battle from PulseData or L&H both of whom lend out there 
> gracious help sparingly.  If we could still interface with the sound 
> hardware, we could use another synthesiser, but I regret that the open 
> source synthesisers available of any real worth (notably Festival and 
> Flight - its derivative) still contain non-free data, though this is a 
> purely philosophical problem.  Anything better, such as IBM viavoice or 
> DECTalk runtime for posix, is not free in the financial sense and/or is 
> outdated or costly or both, despite the outstanding quality.
> 
> > In the linux world right now, there don't
> > seem to be many viable options for using a screen reader with a software
> > synth for speech output. The only options I know of are emacs speak (if you
> > want to call it a screen reader) with I believe the flite speech engine,
> 
> I do not wish to call Emacs a text editor, nor EmacsSpeak a screen reader.  :-)  Do go on...
> 
> > and another package which is called yaser I think, that also uses flite if
> > I'm not mistaken.
> 
> YASR, yet another screen reader.  It works only with EmacsSpeak speech 
> servers for contacting software TTS.  Going through multiple sockets 
> layers does not appeal to me, especially when you consider the already 
> slow speed of Festival, which was meant as a research tool and not as a 
> mainstream synthesiser.  Still, it's a generous offer and could be 
> improved to overcome this problem.  It uses a shell instance to parse 
> console output.
> 
> > Recently however, 2 projects working independently have surfaced, which aim
> > to make it possible to use the speakup screen reader  with the festival tts
> > engine.
> 
> Cute idea, but I'm fed up with kernel-based screen readers.  I offer 
> commendations to those involved and wish to point out that getting it 
> working is likely to take much tallent and alchemy, particularly since the 
> process will probably involve writing some patch-level code for 
> communicating with sound hardware and sockets early in the boot process.  
> However, the feature set and the non-modular nature of Speakup in 
> particular is greating on me a little bit, especially with the recent 1.5 
> release which did not really add anything exciting.  I feel personally 
> that BRLTTY is a very good screen reader, already having TTS support for 
> ViaVoice and Festival; with some review functionality added to synchronise 
> speech and braille output and keyboard navigation which is under 
> consideration it would be ideal for the BN/VN.  Also, let us not forget 
> other efforts, such as Jupiter - another kernel-based screen readers using 
> only a small accessibility patch to communicate with a module, which is 
> very nice.
> 
> > Since speakup is the screen reader that I and many others use,
> 
> I have only the BN attached to my system, which is not supported by 
> Speakup.  Therefore, I use my other computer and a serial link for a 
> console.  I have used Speakup previously with my Apollo and still could 
> but I have found use of terminals (braille or serial) to be really quite 
> convenient of late.
> 
> --snip--
> > Speakup differs from the other screen readers I mentioned, in
> > that it is part of the linux kernel, where as the other 2 are user space
> > programs, one of which relies on the emacs editor. 
> 
> BRLTTy is able to act as a safe replacement for init (8); to me, this is 
> perfectly sufficient for giving exactly the same result.  
> 
> > Also, I believe the braillenote runs on the Arm processor, and I believe
> > someone on the speakup list stated that festival has been ported to the arm
> > architecture.
> 
> The BN/VN uses the Fips processors.  Any idea if Festival (preferably 
> flight) is ported to Fips yet?  Flight, remember, is just a faster 
> Festival.  
> 
> > If someone is interested, I'll dig up the urls for the 2 speakup/festival
> > projects, and send them along, or someone else may do so.
> 
> Please publish those, I'd be interested in the specifics and a look at the code.
> 
> > Well, that's it for now. Thanks for reading.
> 
> Cheers for this.  I'll take a quick moment now to introduce myself as I 
> should have all those years ago... :-)  I'm a repeat fresher (a privileged 
> position indeed - yeah right) at Loughborough University in England, UK, 
> studying computer science.  technically Savvy for a good while, my 
> interests are dynamic, varied and sundry, and while they may not appear 
> convincing on a CV they are definitely not those of a - what do you call 
> them now? - geek, or nerd.  I have no sense of organisation.  I lack time 
> these days, for all the bureaucracy life has to offer, and don't feel 
> particularly like a first year ought.  My clear English, astonishing 
> elloquence and extremely bad sense of humour has always been the source of 
> much criticism, so let the flames begin! :-)  I dunno how best to put me, 
> really.  I'm just this guy, y'know?  It takes getting to know me.  Hope 
> this serves and the discussion picks up.
> 
> > Greg
> 
> Cheers,
> Sabahattin
> 
> -- 
> 
> Thought for the day:
>     Advertising (n): the science of arresting the human
>     intelligence for long enough to get money from it.
>            -- Stephen Leacock.
> 
> Latest PGP Public key?  Click:
> <mailto:PGPPublicKey at sabahattin-gucukoglu.com>
> and send that message as is.
> 
> Sabahattin Gucukoglu
> Phone: +44 (0)20 7,502-1615
> Mobile: +44 (0)7986 053399
> http://www.sabahattin-gucukoglu.com/
> E-mail or MSN Messenger: <mail at Sabahattin-Gucukoglu.com>
> 