Speech Recognition

[Wendy]

Hi all,

I have started tinkering with the built in Windows Speech Recognition within
my Vista operating system. I find it quite novel that you can talk to the
screen and your words come out on it. Granted that I have cheated at times
when it goes wrong and I have to type in corrections but I was wondering if
there are any experts out there who could tell me how to make this more
accurate?

I have also invested in a program called Dragon Naturally Speaking and
loaded it onto my machine today. I started tinkering with this but I
believe I have to learn completely different commands to that of Windows
Speech Recognition.

Does anybody know the difference? I have got the bug with this now and as I
am a teacher, it could come in very handy especially if I could get it to be
more accurate.

Thank you

Wendy

 

[John Barnett MVP]

Wendy you can only really teach it (or should that be 'learn' it - depending
upon whether you listen to Toad or Ratty) by using it on a daily basis
and correcting it with your voice rather than manually typing the
correction, eventually it will get used to your voice pattern and will type
the correct words. I think it is more a question of patience more than
anything else. I must admit Vista's speech recognition is far more intuitive
than that of say XP; and Windows 7 is even better still. I take it that you
have gone through the usual Vista speech recognition tutorial screens? Once
you have done that all you can really do is just talk and correct any
mistakes.

I haven't used Dragon Naturally Speaking since it first came out some years
ago, but i would imagine even that has improved dramatically from the days
when you used to read the tutorial and your voice pattern for each word was
stored in a separate file on your PC. I remember it used to take an age to
safe those voice pattern files


--

--
John Barnett MVP
Windows XP Associate Expert
Windows Desktop Experience

Web: http://www.winuser.co.uk
Web: http://xphelpandsupport.mvps.org
Web: http://vistasupport.mvps.org
Web: http://www.silversurfer-guide.com

The information in this mail/post is supplied "as is". No warranty of any
kind, either expressed or implied, is made in relation to the accuracy,
reliability or content of this mail/post. The Author shall not be liable for
any direct, indirect, incidental or consequential damages arising out of the
use of, or inability to use, information or opinions expressed in this
mail/post..

 

[Wendy]

Thanks for the advice John,

Yes, I have done the tutorial in windows speech recognition and I completed
the same in dragon naturally speaking earlier today but I am finding the
dragon harder to learn because I have just got it.

How this all started was I was tinkering with windows speech recognition and
using it to take short notes for work. I was advised by a friend to invest
in a decent microphone and to try out dragon naturally speaking which I have
since invested in.

Are you saying that it learns as you go along and that you should correct it
by speaking also? You are right about one thing, I have not really got the
time to delve into it in great detail but I do find it intriguing and for
taking notes it is perfect.

I was advised by a good friend to invest in the dragon program because it
was easier to learn but he told me that I would have to learn a completely
different set of commands to that of windows speech recognition and as you
say, to learn it properly would probably take a great deal of time.

Wendy

 

[Charlie Tame]

Wendy you can only really teach it (or should that be 'learn' it -
depending upon whether you listen to Toad or Ratty) by using it on a
daily basis and correcting it with your voice rather than manually
typing the correction, eventually it will get used to your voice pattern
and will type the correct words. I think it is more a question of
patience more than anything else. I must admit Vista's speech
recognition is far more intuitive than that of say XP; and Windows 7 is
even better still. I take it that you have gone through the usual Vista
speech recognition tutorial screens? Once you have done that all you can
really do is just talk and correct any mistakes.

I haven't used Dragon Naturally Speaking since it first came out some
years ago, but i would imagine even that has improved dramatically from
the days when you used to read the tutorial and your voice pattern for
each word was stored in a separate file on your PC. I remember it used
to take an age to safe those voice pattern files

Hehe, despite several attempts for long periods of time each I was never
able to use Vista's speech recognition for anything useful, it was
taking me longer to make the corrections than to type it. Maybe that's
something to do with my UK accent (Simon Cowell ish) but oh dear, some
of the errors it made would have been downright embarrassing had I not
proof read everything very carefully

 

[Charlie Tame]

I have used Dragon Naturally speaking (v9.5 preferred) on an XP machine
in the past with impressive results. This software also has the ability
to transcribe text from recorded files which can be extremely useful,
and this feature appears to be missing from vista's software.

Though I have not taken the time to train vista's engine extensively, I
have been impressed with its "out of the box" accuracy. One of Dragon's
apparent strengths, though, is better correction features which are very
important for "on the job" re-training of the voice models. If the
correction is too time consuming, you won't use it. And if the accuracy
is only 95%, it becomes too time consuming to manually correct the same
errors repeatedly, plus you run the risk of overlooking embarrassing
mis-transcriptions that can be tough to catch without a careful and
focused review that can eat up all the time that dictation should have
saved.

I plan to take the time to really train vista's speech recognition and
I'll post back if it proves to be any better than typing (~50 words/min
for me).

I had surgery a couple of years ago and apparently one side effect can
be a form of dyslexia. For some weeks I couldn't type worth a crap, and
still can't sometimes - it wasn't just missed spelling but also locating
letters on the keyboard. During that time speech recognition should have
been useful, but as you say I spent so much time correcting errors it
really didn't help much, but I had a good incentive to try and make it work.

 

[John Barnett MVP]

Wendy, the whole object of the exercise is for Windows Speech Recognition is
for it to 'learn as it goes along'. It learns by noting what you say and
also what you correct, therefore you should speak the commands to correct
any mistakes. You not only need to get it to recognise the command to delete
one particular word but also selecting sentences etc. I haven't used Vista's
Speech Recognition for some time but in Windows 7 a list of options will
come up. For example, say you want to open Internet Explorer. You would say,
'open Internet Explorer.' If WSR didn't quite understand (as I say this is
Windows 7) a numbered list of options will appear with say 1. windows
explorer; 2 internet explorer; etc and you would then simply say 2 (because
you want to open Internet Explorer) followed by OK. On that command Internet
Explorer will open. Like everything you just need time and patience to learn
the command.

--

--
John Barnett MVP
Windows XP Associate Expert
Windows Desktop Experience

Web: http://www.winuser.co.uk
Web: http://xphelpandsupport.mvps.org
Web: http://vistasupport.mvps.org
Web: http://www.silversurfer-guide.com

The information in this mail/post is supplied "as is". No warranty of any
kind, either expressed or implied, is made in relation to the accuracy,
reliability or content of this mail/post. The Author shall not be liable for
any direct, indirect, incidental or consequential damages arising out of the
use of, or inability to use, information or opinions expressed in this
mail/post..

 

[John Barnett MVP]

Charlie, I too am from the UK and, strangely enough, when I first used
Speech recognition a whole paragraph would be written for me simply because
I 'breathed' on the microphone. I didn't have to say a word Still at
least WSR is more responsive now compared to early versions of speech
recognition. But as you have a voice like Simon Cowell perhaps WSR has taken
a dislike to you, just like a vast majority of viewers have taken a dislike
to the 'real' Simon Cowell

--

--
John Barnett MVP
Windows XP Associate Expert
Windows Desktop Experience

Web: http://www.winuser.co.uk
Web: http://xphelpandsupport.mvps.org
Web: http://vistasupport.mvps.org
Web: http://www.silversurfer-guide.com

The information in this mail/post is supplied "as is". No warranty of any
kind, either expressed or implied, is made in relation to the accuracy,
reliability or content of this mail/post. The Author shall not be liable for
any direct, indirect, incidental or consequential damages arising out of the
use of, or inability to use, information or opinions expressed in this
mail/post..

 

[Patrick Keenan]

Hi all,

I have started tinkering with the built in Windows Speech Recognition
within my Vista operating system. I find it quite novel that you can talk
to the screen and your words come out on it. Granted that I have cheated
at times when it goes wrong and I have to type in corrections but I was
wondering if there are any experts out there who could tell me how to make
this more accurate?

I have also invested in a program called Dragon Naturally Speaking and
loaded it onto my machine today. I started tinkering with this but I
believe I have to learn completely different commands to that of Windows
Speech Recognition.

Does anybody know the difference? I have got the bug with this now and as
I am a teacher, it could come in very handy especially if I could get it
to be more accurate.

Thank you

Wendy

Use the program you're most comfortable with - if you want to use this, you
are going to spend a lot of time with it.

Speech recognition can *never* be 100% accurate, particularly in the English
language (results are better with languages such as Chinese). Homonyms are
a big part of the problem, but there are many other variables such as
inflection, accent, microphone placement, characteristics and quality,
ambient sounds, and dictionaries.

You will *always* have to re-read the entire text, and you will find some
real clangers as you do. This is not in any way "cheating" - it's a basic
necessity.

You must train both yourself and the voice-recognition system, and this will
take time. You must use a decent, directional microphone and turn off or
block other sound sources.

The only context in which speech recognition is remotely reliable is in the
case of a court reporter using a mask. In this case, there is one
highly-trained voice aimed directly at a fixed and contained microphone.
But even then, the text must be re-read and compared to the audio recording.

HTH
-pk

 

[Wendy]

Though I have not taken the time to train vista's engine extensively, I
have been impressed with its "out of the box" accuracy. One of Dragon's
apparent strengths, though, is better correction features which are very
important for "on the job" re-training of the voice models. If the
correction is too time consuming, you won't use it. And if the accuracy
is only 95%, it becomes too time consuming to manually correct the same
errors repeatedly, plus you run the risk of overlooking embarrassing
mis-transcriptions that can be tough to catch without a careful and
focused review that can eat up all the time that dictation should have
saved.

I have also been told the commands and correction features in Dragon are
superior to windows speech recognition but to balance that, the "show
numbers command" in windows speech recognition is a very clever where
everything is on the screen is numbered and you just choose the number of
the item or text you want.

How do you learn all the commands and correction features in Dragon? With
the program I have, it has given me a URL to download acrobat reader and I
gather this shows you all the commands and correction features, is this
right?

Wendy

 

[John Barnett MVP]

Wendy, as with Windows Speech recognition Dragon also has a list of commands
that you can print out. I suggest you download Adobe Acrobat Reader so you
can look at and print the commands. I assume that somewhere on your Dragon
CD there is a pdf file (adobe acrobat) with all the most popular commands
in. All you need do is locate the pdf file, open it in adobe acrobat reader
and checkout the command.

--

--
John Barnett MVP
Windows XP Associate Expert
Windows Desktop Experience

Web: http://www.winuser.co.uk
Web: http://xphelpandsupport.mvps.org
Web: http://vistasupport.mvps.org
Web: http://www.silversurfer-guide.com

The information in this mail/post is supplied "as is". No warranty of any
kind, either expressed or implied, is made in relation to the accuracy,
reliability or content of this mail/post. The Author shall not be liable for
any direct, indirect, incidental or consequential damages arising out of the
use of, or inability to use, information or opinions expressed in this
mail/post..

 

[Charlie Tame]

Yes I think it has improved but I guess I'd need to start over to try it
again and I have not had much time to spend on it.

And no, I don't particularly like Simon either, but I bet he earns more
than I do

 

[Pete]

Hi all,

I have started tinkering with the built in Windows Speech Recognition
within my Vista operating system. I find it quite novel that you can talk
to the screen and your words come out on it. Granted that I have cheated
at times when it goes wrong and I have to type in corrections but I was
wondering if there are any experts out there who could tell me how to make
this more accurate?

I have also invested in a program called Dragon Naturally Speaking and
loaded it onto my machine today. I started tinkering with this but I
believe I have to learn completely different commands to that of Windows
Speech Recognition.

Does anybody know the difference? I have got the bug with this now and as
I am a teacher, it could come in very handy especially if I could get it
to be more accurate.

Thank you

Wendy

I tried Dragon and it was a waste of money. The errors drove me crazy.
This technology is not ready yet IMHO.

 

[Wendy]

I tried Dragon and it was a waste of money. The errors drove me crazy.

This technology is not ready yet IMHO.

Thanks very much Pete, that is a great comfort to me now I have actually
purchased the software.

Yes John you were right, all the instructions, commands and features are
there in a pdf file which I downloaded with Adobe Acrobat Reader and printed
out.

Now I guess it is just a matter of learning the commands etc.

Wendy

 

[John Barnett MVP]

So long as you keep them handy you should soon pick them up. To be honest i
gave up using WSR in Vista because I found, after a while, that I was
getting blinding headaches and I couldn't understand why. I think it was
having to constantly moderate the pitch of my voice so that (a) WSR
recognised what I was saying and (b) the pitch was low enough so as my
family in the next room didn't have to listen to me dictating I found
this to be very stressful.

I've tried WSR in Windows 7 but I had to abandon that because of a bug which
kept switching the microphone on and off - very frustrating when you are
trying to dictate.

With WSR I always kept the list of command on my desk so, anything I was
unsure of, I could quickly locate the command without having to wade through
reams of paper or the WSR help file.


--

--
John Barnett MVP
Windows XP Associate Expert
Windows Desktop Experience

Web: http://www.winuser.co.uk
Web: http://xphelpandsupport.mvps.org
Web: http://vistasupport.mvps.org
Web: http://www.silversurfer-guide.com

The information in this mail/post is supplied "as is". No warranty of any
kind, either expressed or implied, is made in relation to the accuracy,
reliability or content of this mail/post. The Author shall not be liable for
any direct, indirect, incidental or consequential damages arising out of the
use of, or inability to use, information or opinions expressed in this
mail/post..

 

[Charlie Tame]

So long as you keep them handy you should soon pick them up. To be
honest i gave up using WSR in Vista because I found, after a while, that
I was getting blinding headaches and I couldn't understand why. I think
it was having to constantly moderate the pitch of my voice so that (a)
WSR recognised what I was saying and (b) the pitch was low enough so as
my family in the next room didn't have to listen to me dictating I
found this to be very stressful.

I've tried WSR in Windows 7 but I had to abandon that because of a bug
which kept switching the microphone on and off - very frustrating when
you are trying to dictate.

With WSR I always kept the list of command on my desk so, anything I was
unsure of, I could quickly locate the command without having to wade
through reams of paper or the WSR help file.

Ah the files were playing backwards... and you were hearing Bill Gates
saying "Come, join the evil empire" and all the time your real self was
saying no, I want Linux...

 

[Wendy]

May I ask what version of Dragon you bought?

"Standard" or "Preferred" ?

Hi Mark,

It is 9.5 Preferred. I bought it from a friend of a friend for £ 40.00. He
said it was good value and it was still in the box, he had never used it.
The 10 Preferred is too expensive for me, I am only a humble teacher on a
meagre salary, it seemed like a good deal.

I have started to look at some of the commands and features but there are
hundreds! Plus I do get confused and forget this is a different speech
recognition program to windows and occasionally I say the wrong command so I
have to quickly look at my printed out sheet of commands for Dragon to
remind me.

One thing I have noticed with Dragon, as I am dictating the words, there is
a slight delay before the word's appear On The Screen. Do you just carry on
dictating and not worry about this delay? It is very clever though and I am
fascinated with it at the moment. It could be a very useful tool for me if
I had more time. I suppose little by little if I gradually learn the basic
commands at first it will eventually stick in the grey matter!

I am still tinkering after 14 years ;-)

14 years! You must be an expert on the subject, that is a long time and I
can tell by your post that you are very knowledgeable with this technology.

Me, I have just started, four days ago with Dragon! I have tinkered with
windows speech recognition a little more because it is in my operating
system anyway, so I thought I would give it a bash.

I also use a Dictaphone for work to take any last minute notes down and this
technology could also help me in that department albeit just for taking
notes at the moment by dictating.

So nowadays, I have no problem speaking at a fast rate
of 336 wpm and have my computer understand each and
every word.

Gosh, really? That is fast. How long did it take you to dictate that fast?
Do you use it for work in some capacity?

For me, it has the potential for being a permanent tool within my work if I
persevere as you have done. I know where to come if I need any tips then!

Good Luck, Wendy, remember, keep it fun.

Yes, it is sort of at the fun stage at the moment but with a little
frustration added for me in my case.

In your case after 14 years you must know every command and feature like the
back of your hand! Still you have to start somewhere I suppose and if I
study these commands bit by bit every night, I can only improve, right?

Wendy

 

[John Barnett MVP]

Actually Charlie I've tried the latest version of Linux. As always it was
uninstalled just as quickly and consigned to the recycle bin where it
belongs!

--

--
John Barnett MVP
Windows XP Associate Expert
Windows Desktop Experience

Web: http://www.winuser.co.uk
Web: http://xphelpandsupport.mvps.org
Web: http://vistasupport.mvps.org
Web: http://www.silversurfer-guide.com

The information in this mail/post is supplied "as is". No warranty of any
kind, either expressed or implied, is made in relation to the accuracy,
reliability or content of this mail/post. The Author shall not be liable for
any direct, indirect, incidental or consequential damages arising out of the
use of, or inability to use, information or opinions expressed in this
mail/post..

 

[Charlie Tame]

Not sure what you answered there, don't recall mentioning Linux but
never mind. I find that there are similarities with Windows in that a
simple clean system like Debian is very stable while the more you add to
it (Ubuntu for example) the more problems you can get. Obviously that's
partly because you will be doing more with it, but also if you go adding
applications like crazy just because they are there you will get some
bad ones, again just like Windows.

 

[Gene E. Bloch]

Not at all unusual to see a doctor hammering away with a
chisel on a stone tablet here in the good old USA.

You must live in the eastern US. Here in the West we use clay and styli
with wedge-shaped tips. Much more user friendly.

I have enjoyed your encouragement of Wendy. I wish her good luck as well


BTW - I don't use speech-to-text at all, but a friend of mine needs it
because of repetitive motion problems, so I'm interested by proxy, and have
been lurking in this thread.

 

[Charlie Tame]

That is not an accurate statement. WSR (Windows Speech Recognition) was
designed as a general business vocabulary. I use it to respond to dozens of
E-mail a day with very few corrections necessary. Using the WSRTookit I am
able to easily create and use dozens of text macros (repeatable text). In
addition, I have some easily created command macros that scroll from one
customer record to the next. Also I have some sophisticated command macros
for Importing Orders into the Accounting program and Exporting customer
shipping information to our UPS label program.

Of course I do not usually recommend WSR to a physician because medical
terminology and workflow is highly specialized. However, I trained an
orthopedist several months ago on WSR. He dictates only into Microsoft Word
(not an Electronic Medical Record software). We ran 50 of his Word documents
through the WSRToolkit's, "Add Text From File," feature. This parsed his
document and set the WSR Language Model to more closely understand his style
of speaking. See the WSRToolkit at:
http://www.mymsspeech.com/microphones/prod_details.asp?subCatID=71&prodID=228

Marty

But that is not part of Vista so if people don't know about it they will
only see the OS as it stands...