| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1022.1 | English text to phonemes program | YNOTME::WALLACE | | Wed Nov 28 1990 16:28 | 41 |
| One of the difficulties with using a speech chip, is that the chips require
that you send them phonemes (or allophones) as apposed to English (German,
Spanish, etc..) words. Which means that to use a speech chip you really need a
program that will translate words (as well as letters and numbers) into
phonemes.
Aside from the companion translation chips that some vendors provide, there
are two types of software solutions to this problem. One way, is to, by trial
an error create a look up table (or dictionary) of words and there phoneme
eauivelents. The second is to implement some type of algorithm that will
translate any word to it's phoneme equivelents. The first method has the
limitations of having a small subset of words that can be translated, and the
amount of time it takes to figure out the translation for each word used. The
second method has the limitation of only working for a particular language and
of course inaccuracies (inabilities?) of the translation algorithms.
Bryan's current software uses the lookup table approach and has a start on a
dictionary of words and there translations.
In April of 1985, John Wasser posted an implementation of the second approach,
to one of the Usenet news groups. That source code can be obtained from
OLDTMR::$1$DUA8:[WALLACE.PUBLIC.ST]PHONEME.ARC
This English to phoneme translation program, is based on a report circa 1976
by the Naval Research Laboratory. It uses 41 phonemes which are not the same
(at least not the same names) as the ones the SPO256-AL2 chip uses, but I'm
assuming that there is some correlation and that the program can be altered
accordingly. I don't currently have an SPO256-AL2 spec sheet so I won't be
able to make the changes myself.
From the document file (PHONEME.DOC) -
"Here are the programs you need to translate English text to
phonemes. The translation rules are from a government study
and are probably in the public domain. The rest of the code
was written by me and is now in the public domain. Have fun!
-John A. Wasser"
If anyone makes any changes/enhancements to this program make sure you let us
all know.
Ray
|
| 1022.2 | | PNO::SANDERSB | Resist much, Obey little | Wed Nov 28 1990 18:33 | 6 |
|
Any chance of getting a copy of the schematics? Maybe someone
with access to a scanner to create an image file that could be
printed out on a laser printer...
Bob
|
| 1022.3 | Scanner Available | SUBURB::JAMESH | Left Handed People are SUPER Natural | Thu Nov 29 1990 05:08 | 4 |
| I have access to a image scanner and could also probably provide a
Postscript copy of a PC board if anyone is interested.
All I need are the schematics.
...Howard
|
| 1022.4 | Executable program? | MPGS::RADOFF | | Thu Nov 29 1990 15:41 | 12 |
| I am quite familiar with this device having built both hardware and
software. It is very true that the most tedious part is assembling the
phonemes to make a reasonably sounding and understandable speech
generator. I applaud the fact that someone has written some software
that enables this work. It would also be advisable to build a piece of
software that converts the phonemes to the GI hex code since that is
all the the chip understands and must be sent to the device to
announciate the phonemes.
What needs to be done to assemble the source files into an executable?
Steve
|
| 1022.5 | just compile and link all five files | YNOTME::WALLACE | | Thu Nov 29 1990 16:48 | 12 |
| > What needs to be done to assemble the source files into an executable?
Compile the five source files and then compile them with -
$ link parse, phoneme, english, sayord, spellword
You can then run parse and type words at it. I think you can also pass a file
to it on the command line, if you want it to convert a whole file. As output
it just types the phonemes on the terminal.
The code as is not hardware specific so you can compile and run it on either
the VAX or the ST (though I have not tried it on the ST, just on the VAX/VMS).
Ray
|
| 1022.6 | Schematics are already on-line, just a matter of where | PRNSYS::LOMICKAJ | Jeffrey A. Lomicka | Fri Nov 30 1990 12:36 | 4 |
| The schematics were drawn in RAGS on a VAX, and are already in
PostScript form - I will getin touch with Bryan and make them available
here ASAP.
|
| 1022.7 | DEC & Voice | UTROFF::VEENHOF | It's got to be somewhere... | Mon Dec 03 1990 10:33 | 10 |
| For those of you interested in the activities of DEC in speech
technology, it may be worth while to take a look at two conferences:
VOX::VOICE_PRODUCTS (archived, read-only)
VOX::VOICE_PRODUCTS_V2
In these conferences DECtalk and DECvoice are discussed.
Have fun,
Ton
|
| 1022.8 | PS schematic | NOHOST::LEVIN | Bryan, Edu/Comms LKG2-2/U10 | Mon Dec 03 1990 12:21 | 22 |
| I've placed the PostScript-printable file in my atari public area:
NOHOST""::ATARI_VOX:SP0256.PS
(if node NOHOST does not resolve to "5.443" on your system, replace with:
5563""::ATARI_VOX:SP0256.PS
... and tell your NETMGR to update your SYS$SYSTEM:netnode_remote.dat file)
I'll place my initial pass at a dictionary here, as well as other
source files when I get time.
BTW, I drew the art for this, so if you need it/want it (for revisions,
etc), please contact me.
Cheers,
.bl
|