Print

Print


>I understand from reading your recent messages you are
using  Dragon
>Systems software.  I have been considering it myself.  I
have a pentium
>166 with lots of RAM and a reasonably fast hard drive.  I'm
curious about
>what you meant about processing speed increased. Does that
mean you could
>talk faster or what?  What does the software do when you
overrun it.  I
>have never tried any voice recognition software, so I would
appreciate
>your information.


For Julian and others who have written about
voice-recognition software.

This type of software serves two purposes
1. As a taker of dictation for entering into a word
processor or editor.  One has the WP program and the VR
program running at the same time.
2. As a means of navigating the screen (UP, DOWN, NEXT PAGE,
etc.), and issuing menu commands when in a compatible
program (FILE OPEN, HELP, FILE PRINT, etc.)

So when you are using it, you will always be running at
least two (usually large) programs at once.  This is one of
the reasons why such programs require lots of RAM memory, at
least 32Mb.

The other drag on the system (is that why one of them is
called 'Dragon'? <grin>) is the way the program works.

When you speak a word into the microphone, the program makes
a frequency 'map' in RAM memory.  It then looks through all
the standard maps it knows, to find one which is as close a
match as possible to the one you just spoke.  Of course, all
this takes millionths of a second, but means that as well as
the two programs being in memory, you now have the two
'maps' in memory as well!

The other requirement to run voice recognition software is a
fast CPU.  This must be a Pentium (for a PC; suitable
equivalent for a Mac, etc.), at least 166Mhz.  Although it
is *possible* to run such a program with a slower processor,
you will go 'nuts' waiting for it to do something.  The more
RAM you have, and the faster the CPU, the faster you can
dictate.  If one were only going to use the program for
navigating and issuing commands, in other words as a
replacement for a keyboard and/or mouse, the speed of the
processor could be slower.

The other major consideration with VR software is that it
has to be 'trained' to recognise *your* particular voice.
The original software has a 'dictionary' of words and
phrases, anywhere from 60,000 to 100,000, depending on
brand. BUT... they are in a standardised form of the
language wanted, eg, "UK English", "US English", etc.  If
you don't speak in this 'standard' way, the program will not
'understand' you.  So if you were, for example, an American
from the 'deep South', or had a western drawl, or have a
foreign accent, eg, French, Dutch, etc., you will have to
generate your own 'standard' language 'map'.  The more you
differ from the norm, the more work you will have to do.
This applies to people with speech difficulties as well.
Then, because your customised 'map' may be larger than the
standard 'map', you will need further RAM.

As well as that, you will need to 'train' the program to
learn words, phrases, spellings, etc., that are additional
to the standard 'dictionary', for example, names of
companies, unusual spellings of first names, medical terms,
etc.  It is possible, mostly, to buy extra custom
'dictionaries' such as Medical, Legal, Architectural, etc.
and/or to have these made for you.

So don't expect a miracle program, and be prepared to do a
lot of setting-up work before you can sit back at ease and
command your PC by voice.

Minimum workable requirements for a PC would be:
32Mb Ram, Pentium 166Mhz CPU, approx 20Mb spare space on
your Hard Disk, plus 50-100Mb of space for the program.

BTW, if a person has great difficulty speaking, such as with
severe cerebral palsy, it will be almost impossible to
'train' the program; apart from anything else, many of the
person's word and phrase soundings will be so similar that
the program will not be able to differentiate between them,
and will generate errors, which then have to be corrected.

Further to that, if the person is intellectually handicapped
as well, they may not be able to understand the strict
guides within which the program operates, and so will not be
able to use it successfully.  In either of these cases,
alternate software/hardware may be necessary.

To sum up, the people who may need this type of program
most, the less-well-off and more-disabled, are the people
who will benefit from it least.  Its main use is as a
business aid, and that will be so until super-fast
software/hardware and very cheap RAM memory become
available.

Jim Slattery, AACS  59/13
[log in to unmask]