Yes I can also confirm all of that. I never dictate, only eve r use the echo
for commands and simple stuff.
The Apple dictate a text can leave one amused.
Take, feeling better, but leftme with a nasty chesty cough Actually came
out in text as left me with a nasty testicle.
So be careful out there!
Brian

--
----- -
This newsgroup posting comes to you directly from...
The Sofa of Brian Gaff...

Blind user, so no pictures please!
"The Natural Philosopher" wrote in message
news
On 11/02/18 11:58, Peter Parry wrote:
On Sun, 11 Feb 2018 10:10:26 +0000, The Natural Philosopher
wrote:



Well yes, IO was after feedback from people who had iused it
extensively, to write books...or prepare lengthy reports.


I used Dragon Naturally Speaking about 5 years ago and the system
built into Google Docs/Chrome ( Voice Typing) more recently.Dragon was
to try to use for producing procedure manuals which were quite
lengthy. Voice Typing was for regular non technical text.

Voice recognition is something which is very processor intensive but
it hasn't really advanced as much as everyone thought it would despite
processor performance improving.

On all the systems I have tried one universal requirement is for a
very good microphone and careful positioning. The best was a passive
noise canceling microphone from a Nokia car kit held about one inch to
the side of my mouth so no impulse noise. The microphone was used
with a home made "Madonna microphone" band to keep the microphone
secure and stable. The Nokia microphone is a simple small one but
optimised for speech and good at noise canceling (it came from the
days when Nokia knew what they were doing). I think I still have a
few left over so if you want one let me know.

As for the software Dragon worked quite well and benefited
considerably from time spent training it to your voice. A disabled
colleague who used it all day both for control and writing had a very
low error rate.

Google Voice Typing I've used less but it wasn't bad (and free), it
can format text as well as recognising speech and it is easy to take
text from Google Docs into Word if you want.

Problems with both (indeed all voice recognition software) was that
they never made mistakes, or at least they never admitted it. If you
said "Tomorrow is (mumble) Thursday" and they didn't quite get it then
rather than saying "yer what?" they would find the closest word so
"Tomorrow is Teatime".

Unlike keyboard entry you can't pick up these mistakes with a spell
checker - nothing gets spelled incorrectly by speech recognition
software, its words which get changed. If you are a good proof reader
you can pick these errors up. If like me you are a lousy proof reader
- seeing what I think should be there rather than what is there - it
is a pain and the main reason I don't use voice recognition.

It can also be serious if you are writing something like a manual
where accuracy is essential.

Technical words were a particular problem and at the beginning a lot
of time had to be spent teaching Dragon new words. To be fair its
built in vocabulary is very good for normal documents but things like
Thyristor, Fast Fourier etc tended to give it indigestion.

Dragon needed a large investment of time at the beginning both to
enter technical words and to get the voice training and microphone
positioning optimised. Once you have made this investment in time it
was good. It was also expensive.

Voice Typing was for a seriously ill person so there wasn't time to do
anything more than get the microphone right and the application
working. It was being used for normal (non-technical) correspondence,
did a competent job and allowed them to write what they wanted. During
testing I was impressed by how well it worked especially for a free
application. It would be worth trying (as it is free).

With all voice recognition acting on commands is far more accurate
than handling free text. With only a handful of words to recognise
command recognition is near perfect. Free text I never measured the
error rate accurately but in a quiet room (they don't like background
noise and fan noise in particular) it was about one error per 200
words. As I said though it wasn't always easy to pick out the mistake
when it took place.

Thank you Peter, that is most valuable feedback


--
Microsoft : the best reason to go to Linux that ever existed.