home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
EnigmA Amiga Run 1996 February
/
EnigmA AMIGA RUN 04 (1996)(G.R. Edizioni)(IT)[!][issue 1996-02][Skylink CD III].iso
/
earcd
/
library
/
spnshccn.lha
/
spanish_accent
/
Programmers.readme
< prev
next >
Wrap
Text File
|
1995-10-08
|
4KB
|
99 lines
This file contains information included in the guide file in spanish that
is intended to be read by people willing to improve the
translator-narrator-speak system and specially by Francesco Devitt
PROBLEMS WITH TRANSLATOR.LIBRARY:
- Español.accent can't stress correctly words with stress on the last
syllabe without a written tilde (´) like "tapiz" that has to be writen
"tapíz" to be correctly spoken.
The translation rules defined by Francesco Devitt for the accent files
doesn't allow Amiga to know when a word without tilde is stressed on the
last syllabe or the syllabe before the last.
We know it because if it is stressed on the syllabe before the last, it
ends on vowel, n or s.
the problem is that there is no way to indicate translator library to
find if a word ends on one of this letters.
Anyway the more i think about it, the more i believe there is a way of
doing this using the existing rules but i can't find it :-(
- Español.accent can't translate correctly numbers greater than 9999.
The translation rules defined by Francesco Devitt for the accent files
doesn't allow Amiga to translate correctly an undefined sucession of digits.
Actual implementation uses a different rule for every position of the same
number at the numeric string, e. g.: this are the rules for number 4:
[40](dg~) = { cuarenta } /* forty */
[4](dg)(dg~) = { cuarenta y } /* forty something */
[4](dg)(dg)(dg~) = {cuatrocientos} /* four hundred */
[4](dg)(dg)(dg)(dg~) = { cuatromil } /* four thousands */
[4] = { cuatro } /* four */
The third rule says: if you find a 4 followed by 2 more digits and
something that is not a digit, translate it with "cuatrocientos".
Is easy to deduct that for every new digit we want to add to the recognition
ability it will need a new rule for every number, this means 9 more rules
for every power of ten.
If Francesco Devitt changes the rules it would be easy to increase the
ability with only a rule for every power of 1000!
if we look carefully how are spoken the numbers in spanish (and similarly
in every other language), we notice that they are divided in groups of 3
digits separated by the words "mil", "millones", "mil", "billones", "mil",
"trillones", etc.
E. g. if we take number 123123 (ciento veintitrés mil ciento veintitrés or
one hundred and twenty three thousands one hundred and twenty three), we can
see that the first 3 digits are spoken like the last 3 with the word "mil" or
"thousands" separating them.
We could define the same rules currently defined adding that they can be
followed by any amount of 3 digit groups (including no one) and define
other rules for generating the words "mil" (thousands), "millón" (millions),
etc.
The actual rules allow to define 3 digit groups but they don't allow to
substitute the empty string by a word that is what is needed for the rules
of "mil", "millon", etc.
PROBLEMS WITH NARRATOR.DEVICE:
This device has information about every phoneme in the American english so
it can speak it correctly.
But almost all other existing languages (including non american english)
have phonemes that are not defined in narrator.device.
Nowadays the only solution is to use the american phonemes that best suit
the sound we want to produce, this inevitably makes Amiga speak like a
cowboy in any other language.
The best solution would be if somebody modifies or creates a new
narrator.device that knows about phonemes of several languages.
Also, translator.library has to be modified to include definitions for the
new phonemes.
PROBLEM WITH SPEAK.HANDLER:
There is a problem with SPEAK-HANDLER v37.4. It seems to filter extended
ASCII so words like "cigüeña" (stork in spanish) or "heiß" (hot in german)
are transformed into "cig e a" and "hei " eliminating characters greater than
127 before sending the string to translator library.
The best solution woul be if somebody patches or creates a new speak.handler.
If somebody fixes any of this problems, please contact me or update this
AmiNet archive. |
|
Juan Carlos Pérez Walls <-----------------------------'
C/ Eliseo López #1-11-A
38010 Santa Cruz de Tenerife
Canary Islands, SPAIN
E-mail:
Sorry, no e-mail, yet...
Telephone:
SPAIN + 922-64-03-67