Prev: Flash and Safari 4.0.4
Next: Successful Speech Recognition with loud outside noise
From: Mark Conrad on 22 Nov 2009 04:14

Okay, Steve Jobs and I are on the same page
regarding this speech recognition (SR) issue.

We both realize that SR is good for Apple, good
for Mac users, good for business users of Macs,
and _very_ _good_ for Windows users who
want to get away from the confusing obfuscated
virus-ridden Windows OS, and switch to
easy-to-use Macs.

....all without losing excellent SR productivity
capabilities, thanks to a tiny company
named MacSpeech.

Why should Windows users be the only ones to
benefit from a modern _free_ SR app built into
their OS, a darn good one at that, every bit
as good as MacSpeech, which we Mac users
have to shell out an extra $200 for.

That is what Steve Jobs was thinking of
ten years ago, when he tried to get SR
for the Mac OS.

<http://www.voicerecognition.com/news/05_10_99.html>


At the last moment, the deal fell through,
because the PC crowd decided that the Mac
market was too small to mess with, so they
decided to concentrate exclusively on the
PC/Windows market, so far as SR was concerned.


I think almost everyone here knows that SR has
limited uses, but the same can be said of
every other type of app for the Mac.

They ALL have limited uses, so what?

I suggest that most Mac users (and PC users)
only use one percent or less of the capabilities
of their personal computers.

That is the beauty of modern personal computers,
we can do thousands of different things with them,
depending on our "personal" preferences.


'Nuff of the pep talk for those Mac users without
imagination and foresight, right now I have bigger
fish to fry.

As you know, Google is NOT my friend.

So I will be absent from these Mac NGs for a
period of time, doing some research on what the
MacSpeech software is actually doing, as contrasted
against what it is _supposed_ to be doing.

Primarily the slider adjustment in MacSpeech
preferences, which appears to have no effect
whatever, whether it is adjusted for maximum
accuracy, or maximum speed.

In theory, when that slider is adjusted for
maximum speed, the text accuracy should suffer.

Text accuracy does not suffer, so I have to mount
a research effort to determine _why_.

It involves firing up my $1,600 Dragon SR app,
to verify that the same slider _does_ work
in that particular SR app.


I know, I know, there will be a lot of sobbing
and crying in these Mac NGs, when they do not get
their daily doses of SR trivia from me.

Try to bear up in the face of this adversity,
with any luck I should be back in about a week.

Mark-
From: Mark Conrad on 24 Nov 2009 00:45
In article <221120090114043520%aeiou(a)mostly.invalid>, Mark Conrad
<aeiou(a)mostly.invalid> wrote:

> I know, I know, there will be a lot of sobbing
> and crying in these Mac NGs, when they do not get
> their daily doses of SR trivia from me.
>
> Try to bear up in the face of this adversity,
> with any luck I should be back in about a week.

Ha, that was a quick week!

I found out what I wanted to find out, about the
Accuracy/Speed slider adjustment in MacSpeech.

Believe it or not, Google was actually a tiny bit of help,
miracle of miracles.

Speech Recognition (SR) newbies look at that slider
adjustment, and do not know how to set it.

Then they complain when their text accuracy goes
down the drain.

FWIW, when I set it, I crank it all the way to the left,
for maximum speed.

Advice from SR experts in the various SR forums is to
set it half way between the speed and accuracy extremes.

That works pretty well, however...


What I found out by experimenting, with both MacSpeech
and Dragon, is to *AVOID* setting it all the way to the right,
the maximum accuracy position.

If you do, all hell will usually break loose, unless your Mac
is a fast supercomputer.

Seems that something inside the SR apps starts gobbling up
CPU cycles, and the net result is that your text accuracy
goes w-a-y down, because your original voice audio signal
gets "chopped up" before it can be processed into text.

You can often hear this chopped-up audio when you activate
"playback" of some long passage you have just recently
dictated.

Chopped-up audio results in terrible speech recognition.


The SR experts on the web have a variety of conflicting advice,
depending on which expert you listen to.

They mostly all suggest initially placing the slider halfway
between the speed and accuracy extremes.

Then play with adjusting it, if you are so inclined.

I just throw it all the way back to maximum speed, and I do not
seem to lose any detectable accuracy - - - YMMV.

I find that _many_ things affect text accuracy much more
than that slider does.

Mark-
 | 
Pages: 1
Prev: Flash and Safari 4.0.4
Next: Successful Speech Recognition with loud outside noise