[MUD-Dev] Speech to Text, etc. (was: On socialization and convenience )

[Travis Casey]

On Monday June 25, 2001 13:01, Madrona Tree wrote:
>> From: "Adam Martin" <amsm2 at cam.ac.uk>

>> FYI I very much doubt that STT for eastern languages
>> (e.g. Mandarin) is any harder than for western
>> languages. Certainly inflection has a strong effect on meaning,
>> but the work I've seen so far seems to indicate that that is not
>> where the real difficulty lies anyway - the problem is more that
>> with sloppy enunciation multiple words become almost
>> indistinguishable (the system I refer to earlier claimed to be
>> able to uniquely decode such situations more easily than current
>> systems do).

> Text-to-speech would be fabulous, but I think speech-to-text would
> be terrible.  Read depositions sometimes - the amount of ums and
> uhs and hmms will amaze you.  I think of the time my father said,
> during a deposition, when someone was pushing records cart down
> the hallway, "Holy S---!  Are we having an earthquake?!"  and the
> court reporter kept on typing.

Or, for something more easily accessible to many people, read the
transcripts of the Nixon tapes.  Those are especially interesting in
that they record the speech of highly educated, intelligent people
-- and yet, they're still look nothing like what most people *think*
speech would look like.

> Come to think of it, though.. maybe seeing what we are actually
> sounding like would be *good* for us.  :P The spoken word so
> easily comes and goes, but the written usually makes a much bigger
> impression.  What a tool for grade-schoolers...!

> But really - good in theory, but I don't think it would work too
> well for our application - especially when many of these muds are
> in pseudo-medieval scenarios... can you imagine yourself having to
> actually *say* thee and thou, without an um in between?  :)

Why would you want to?  As I've said many times before, unless
you're trying to have your character speak an archaic dialect of a
language in the game, there's no reason why you should use
archaic-sounding speech.  The character's own language is not going
to sound archaic to them.

Some personal notes on using speech-to-text programs: my wife and I
practice attachment parenting with our daughter.  As part of that,
when she was a baby, we always kept her with one of us.  Elisa (my
wife) usually used a baby sling, but I never got comfortable with
one, and generally held Rhianna (my daughter) in one arm.

As a result, typing at a rate anything like what I'm used to was
impossible.  So, I picked up a copy of Dragon NaturallySpeaking and
tried using it for a while.  During the course of that, I even
played a couple of muds with it.

Learning to pause instead of saying, "um..." or "uh..." wasn't very
hard.  However, doing it very much created problems, since DNS works
better with continuous speech than with individual words.  For me,
conversations on muds were easier than doing commands, because I'd
generally be able to create a longer stream of words without a pause
for thought when conversing.

DNS had key phrases to tell the system to stop transcribing and to
start again.  It also had a few meta-keys that could be pressed to
give additional information to the system.  For example, you could
spell words out using the alpha-bravo-charlie alphabet, but doing so
was helped by pressing a key to let the system know that was what
you were doing, so it wouldn't try to put down the words instead.
There was a key to press to let the system know you wanted to put
down something as numerals -- e.g., as 333 instead of three hundred
thirty-three.  In DNS, those were a bit difficult to use because
they were key combinations, chosen to make them unlikely to conflict
with a program's shortcut keys.  In an application that had its own
speech-to-text system, though, you could assign them to easier key
combinations.

After Rhi got bigger, I gave up using DNS.  It got fairly good at
recognizing my speech, but it was never as fast as typing for me.  I
have a friend, though, who is developing carpal tunnel to whom I
gave my copy, and he's currently using it and finding it very nice.

--
       |\      _,,,---,,_     Travis S. Casey  <efindel at earthlink.net>
 ZZzz  /,`.-'`'    -.  ;-;;,_   No one agrees with me.  Not even me.
      |,4-  ) )-,_..;\ (  `'-' 
     '---''(_/--'  `-'\_) 

_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev