asterisk users - Dec 2007 - Sounds transscript / speech synthesis

Hi,

in the earlier version there was a sounds.txt with the transcript of the
soundfiles. Does this still exist somewhere?

Is there a plan to make speech synthesis available the same way as
soundfiles, ie. instead of playing language/soundfile.wav, send the text to
the speechengine and play the output...?

Jay...
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.digium.com/pipermail/asterisk-users/attachments/20071222/89e6a485/attachment.htm

On Sat, Dec 22, 2007 at 12:55:19PM +0100, Jay R. Worthington
wrote:
> Hi,
> 
> in the earlier version there was a sounds.txt with the transcript of the
> soundfiles. Does this still exist somewhere?
Not an up to date or complete one, the last time I looked, so I ended up 
transcribing them all so that we could re-record them into UK English.

We adapted a few of the phrases so as not to confuse UK users. Instead of 
"pound" we say "hash" and instead of "Password" we
say "PIN" as some users
had no idea what was meant by "hash" or "password", and
since they have SIP
passwords and login passwords, this was just another "password" we
could do
without confusing them with!

I can send you a copy of the document if you want. There are over 300 
individual sounds with a few extra that we added.

Even then I've had complaints from users that the voicemail menus are too 
long-winded and although functional, it isn't the easiest of systems to use 
compared to others. (We re-recorded the vm prompts pretty much word for 
word, but shortened a few of them, so for example:

"To exit voicemail press the pound key." ended up as:
"To exit, press hash."

The next time we do any recordings, we may record it in a better style, 
as the default sounds are considered bad practice by some(?):

"Press 1 to record your unavailable message. Press 2 to record your busy 
message. Press 3 to record your name. Press 4 to record your temporary 
greeting.  Press 5 to change your PIN.  Press star to return to the main 
menu."

I want to re-record it so that "press x" is after the option, so when
you
hear the option you want, if you were not paying attention, you don't miss 
the number you were supposed to press, and you don't have to listen to all 
the options all over again, so:

"There are five options. To record your unavailable message, press 1. 
To record your busy message, press 2. To record your name, press 3. to 
record your temporary greeting, press 4. To change your PIN, press 5. To 
return to the main menu, press star."

Experience shows that users seem to prefer it this way around. There are 
some inconsistencies between the various sound sets, for example, some 
menus, "press star for help" implies that there is further help. What
they
actually mean is "to hear the menu again, press star" In other menus,
* goes
back up to the previous menu, or does nothing. In some menus we get told to 
"dial" and others "press". Some options require pressing the
# key after,
and others not.

We also edited all the embarrassing "I am soooooorrrrrry..." type 
announcements, or removed "I" whenever it appears."

Machines cannot relate to callers as "I". It's all a bit fake and
insincere.
The machine cannot be "sorry" Callers here find these sort of
"I'm not
really sorry, I'm a machine" announcements annoying and somewhat 
patronizing.

(Using a recorded announcement to try and get 'personal' and
'friendly' just
doesn't work; it's clearly a recording and couldn't care about your
feelings
one way or another, so don't insult my intelligence by pretending it does.)

Here in lovely England, an American woman saying "I am
soooooorrrrrry.."
provokes anger and the response along the lines of "*STOP TELLING ME
YOU'RE
SORRY AND JUST F*((*! DO IT!" followed by slamming the phone down/pounding 
of fists on the desk! I kid you not. It's probably just a cultural thing, 
but I wanted callers not to get annoyed by the system, so we embarked on the 
mammoth 5 hour task of re-recording every single sound, as the small group 
of pilot users we had before the main system went live, all hated the 
default sounds!
> Is there a plan to make speech synthesis available the same way as
> soundfiles, ie. instead of playing language/soundfile.wav, send the text to
> the speechengine and play the output...?
I've heard one or two systems using this, but it sounds a bit strange in my 
opinion, and sometimes rather difficult to hear.

I don't know if there are any plans to change the way asterisk plays sounds 
(I imagine getting it to say everything correctly in many languages would be 
a long and complex challenge!) I'd sooner spend some time recording the 
sounds rather than spend ages listening to the output of the text to speech 
system and tweaking phonetics to get the intonations/accent right.


Rob


-- 
Robert Lister     - London Internet Exchange - http://www.linx.net/
sip:robl at linx.net - inoc-dba:5459*710        - tel: +44 (0)20 7645 3510