«

Hello.

I have Dragon Naturally Speaking but have a little problem.

I have set myself up as a user without any trouble. However I also
wish to have a transcription from a recorded sound file on my computer
of an American Talk Show host.

Is there anyway I can train the system to recognise his accent etc of
this host and transcribe accurately?

When you first set up a new user there is an option to record the
words from a wav file as your dictation source, however, you have to
repeat a certain story back to the system which is already in there.
E.g. 3001 The Final Odyssey, Alices Adventures in Wonder Land, Charlie
and the Chocolate Factory etc...

This would be a problem because I would have to find all the
individual words the Talk Show host says to put them all together for
the story for that to work, and there are quite a few paragraphs of
text for those stories.

Is there any way I can import my own text into the system and train
from that? This way, I would be able to type up a couple of
paragraphs that the Talk Show host actually says in the wav file, and
then I would simply be able to press play and it would all work.

If it is possible to do that please let me know. Otherwise I may have
to just base it on my own voice, and then there will be lots of
inaccuracies as I am British.

Thanks for any help on this.

John

I can't help you and I am not even sure anyone can but I have a question for
you. How well does it do at recognizing your voice?
I bought Via Voice about three years back and it was horrible. I spent
the better portion of a day reading to my computer to train the program and
it still stunk. I tried to dictate one email with it and gave up as even
though I was a slow typist I could still just type faster than I could
correct all the mistakes the program made. It got the wrong word altogether
in many cases and rarely picked the right one out of the same word (to, too,
or two / their, there, or they're).
Did you make your post here with it if so how many mistake were there or
how many mistakes would there have been if you had? If these products have
gotten better I would not mind trying one again but the one I tried was a
waste of money.
Joe
"John" <> wrote in message
news:...
> Hello.
>
> I have Dragon Naturally Speaking but have a little problem.
>
> I have set myself up as a user without any trouble. However I also
> wish to have a transcription from a recorded sound file on my computer
> of an American Talk Show host.
>
> Is there anyway I can train the system to recognise his accent etc of
> this host and transcribe accurately?
>
> When you first set up a new user there is an option to record the
> words from a wav file as your dictation source, however, you have to
> repeat a certain story back to the system which is already in there.
> E.g. 3001 The Final Odyssey, Alices Adventures in Wonder Land, Charlie
> and the Chocolate Factory etc...
>
> This would be a problem because I would have to find all the
> individual words the Talk Show host says to put them all together for
> the story for that to work, and there are quite a few paragraphs of
> text for those stories.
>
> Is there any way I can import my own text into the system and train
> from that? This way, I would be able to type up a couple of
> paragraphs that the Talk Show host actually says in the wav file, and
> then I would simply be able to press play and it would all work.
>
> If it is possible to do that please let me know. Otherwise I may have
> to just base it on my own voice, and then there will be lots of
> inaccuracies as I am British.
>
> Thanks for any help on this.
>
> John
>
>

Joe wrote:
>
> I can't help you and I am not even sure anyone can but I have a question for
> you. How well does it do at recognizing your voice?
> I bought Via Voice about three years back and it was horrible. I spent
> the better portion of a day reading to my computer to train the program and
> it still stunk. I tried to dictate one email with it and gave up as even
> though I was a slow typist I could still just type faster than I could
> correct all the mistakes the program made. It got the wrong word altogether
> in many cases and rarely picked the right one out of the same word (to, too,
> or two / their, there, or they're).
> Did you make your post here with it if so how many mistake were there or
> how many mistakes would there have been if you had? If these products have
> gotten better I would not mind trying one again but the one I tried was a
> waste of money.
> Joe
> "John" <> wrote in message
> news:...
> > Hello.
> >
> > I have Dragon Naturally Speaking but have a little problem.
> >
> > I have set myself up as a user without any trouble. However I also
> > wish to have a transcription from a recorded sound file on my computer
> > of an American Talk Show host.
> >
> > Is there anyway I can train the system to recognise his accent etc of
> > this host and transcribe accurately?
> >
> > When you first set up a new user there is an option to record the
> > words from a wav file as your dictation source, however, you have to
> > repeat a certain story back to the system which is already in there.
> > E.g. 3001 The Final Odyssey, Alices Adventures in Wonder Land, Charlie
> > and the Chocolate Factory etc...
> >
> > This would be a problem because I would have to find all the
> > individual words the Talk Show host says to put them all together for
> > the story for that to work, and there are quite a few paragraphs of
> > text for those stories.
> >
> > Is there any way I can import my own text into the system and train
> > from that? This way, I would be able to type up a couple of
> > paragraphs that the Talk Show host actually says in the wav file, and
> > then I would simply be able to press play and it would all work.
> >
> > If it is possible to do that please let me know. Otherwise I may have
> > to just base it on my own voice, and then there will be lots of
> > inaccuracies as I am British.
> >
> > Thanks for any help on this.
> >
> > John
> >
> >

As i understand it, there are 2 types of speech to text programs.
One requires training, and "recognizes" only that person.
The other might be considered "generic" in that no training is
required, and "recognizes" "anyone" within certain limitations.
I have heard that neither type (for the PC) has ever been reliable or
routinely useable.
The qualification is on the second type; it works as long as the
requirements are met.
For example, you may have called a few companies, including the loco
(pun intended) version of Ma Bell and were prompted to speak answers.
In those cases, the responses (in each case) were rather limited, and
counting of pauses (in the word) and consonants could be used (ie:
ignore the words, per se).
For the PC version, the "vocabulary" starts undefined and one must
speak in a rather formal and precise manner (that is what i heard,
anyway).

Dragon is as good as it gets and works very well (after training.)

Robert Baer wrote:
>
> Joe wrote:
> >
> > I can't help you and I am not even sure anyone can but I have a question for
> > you. How well does it do at recognizing your voice?
> > I bought Via Voice about three years back and it was horrible. I spent
> > the better portion of a day reading to my computer to train the program and
> > it still stunk. I tried to dictate one email with it and gave up as even
> > though I was a slow typist I could still just type faster than I could
> > correct all the mistakes the program made. It got the wrong word altogether
> > in many cases and rarely picked the right one out of the same word (to, too,
> > or two / their, there, or they're).
> > Did you make your post here with it if so how many mistake were there or
> > how many mistakes would there have been if you had? If these products have
> > gotten better I would not mind trying one again but the one I tried was a
> > waste of money.
> > Joe
> > "John" <> wrote in message
> > news:...
> > > Hello.
> > >
> > > I have Dragon Naturally Speaking but have a little problem.
> > >
> > > I have set myself up as a user without any trouble. However I also
> > > wish to have a transcription from a recorded sound file on my computer
> > > of an American Talk Show host.
> > >
> > > Is there anyway I can train the system to recognise his accent etc of
> > > this host and transcribe accurately?
> > >
> > > When you first set up a new user there is an option to record the
> > > words from a wav file as your dictation source, however, you have to
> > > repeat a certain story back to the system which is already in there.
> > > E.g. 3001 The Final Odyssey, Alices Adventures in Wonder Land, Charlie
> > > and the Chocolate Factory etc...
> > >
> > > This would be a problem because I would have to find all the
> > > individual words the Talk Show host says to put them all together for
> > > the story for that to work, and there are quite a few paragraphs of
> > > text for those stories.
> > >
> > > Is there any way I can import my own text into the system and train
> > > from that? This way, I would be able to type up a couple of
> > > paragraphs that the Talk Show host actually says in the wav file, and
> > > then I would simply be able to press play and it would all work.
> > >
> > > If it is possible to do that please let me know. Otherwise I may have
> > > to just base it on my own voice, and then there will be lots of
> > > inaccuracies as I am British.
> > >
> > > Thanks for any help on this.
> > >
> > > John
> > >
> > >
>
> As i understand it, there are 2 types of speech to text programs.
> One requires training, and "recognizes" only that person.
> The other might be considered "generic" in that no training is
> required, and "recognizes" "anyone" within certain limitations.
> I have heard that neither type (for the PC) has ever been reliable or
> routinely useable.
> The qualification is on the second type; it works as long as the
> requirements are met.
> For example, you may have called a few companies, including the loco
> (pun intended) version of Ma Bell and were prompted to speak answers.
> In those cases, the responses (in each case) were rather limited, and
> counting of pauses (in the word) and consonants could be used (ie:
> ignore the words, per se).
> For the PC version, the "vocabulary" starts undefined and one must
> speak in a rather formal and precise manner (that is what i heard,
> anyway).

On Thu, 11 Nov 2004 20:06:47 +0000, John <> wrote:

>Hello.
>
>I have Dragon Naturally Speaking but have a little problem.
>
>I have set myself up as a user without any trouble. However I also
>wish to have a transcription from a recorded sound file on my computer
>of an American Talk Show host.
>
>Is there anyway I can train the system to recognise his accent etc of
>this host and transcribe accurately?
>
>When you first set up a new user there is an option to record the
>words from a wav file as your dictation source, however, you have to
>repeat a certain story back to the system which is already in there.
>E.g. 3001 The Final Odyssey, Alices Adventures in Wonder Land, Charlie
>and the Chocolate Factory etc...
>
>This would be a problem because I would have to find all the
>individual words the Talk Show host says to put them all together for
>the story for that to work, and there are quite a few paragraphs of
>text for those stories.
>
>Is there any way I can import my own text into the system and train
>from that? This way, I would be able to type up a couple of
>paragraphs that the Talk Show host actually says in the wav file, and
>then I would simply be able to press play and it would all work.
>
>If it is possible to do that please let me know. Otherwise I may have
>to just base it on my own voice, and then there will be lots of
>inaccuracies as I am British.
>
>Thanks for any help on this.
>
>John
>

I have upgraded every version since the first in search of the Holy
Grail and despite those who proclaim that it has "arrived", it has
in fact not arrived and I suspect the newest version is no different.
I have trained and trained and trained, bought high priced mics, etc
etc and it still types "for" when I mean "fore" (as just one of a
multitude of examples. This is difficult technology.

It appears that Wizard disagrees with this opinion but believe me, as
a lawyer a good VR system would be a Godsend and it just ain't there
yet. I figure 5 more years...

MA Bell's voive recogition system costs at least $60,000.
Voice recognition has arrived... for $60,000 to $120,000.
How much did you all say you spent?

But all seriousness aside....
Have you tried patching an inexpensive graphic-equilizer between the
source (voice, wav file player) and the input port of your computer?
The idea is to create a "tone" Dragon can better decipher. This is
not the end-all solution. By tone I mean consonants that are
distinctive and that really "pop". As a starting setting on your
graphic-equilizer raise the equalizer's treble and mid range knobs as
high as they can go and set bass knob at zero or below zero. Adjust
them until Dragon gives the best results.

For the radio guy if you don't want to buy an equilizer just adjust
the treble, bass and mid-range knobs on your radio before you record.
If you already have a wav file then adjust tone settings (treble, mid
and bass) on the equalizer then make a dupe.

Summarizing, the voice setup goes in the direction of the arrow.

======>
Voice to mike to graphic-equalizer to computer-input-jack to Dragon.

The radio setup goes in the direction of the arrow.
Radio to equalizer to wave-file-recorder to input-jack to Dragon.

On Sun, 14 Nov 2004 18:24:52 GMT, (gg) wrote:

>MA Bell's voive recogition system costs at least $60,000.
>Voice recognition has arrived... for $60,000 to $120,000.
>How much did you all say you spent?

Now I believe that to be true. This is why I think we are still at
least 5 years out. If the technology really worked now everyone would
have it and use it. The reason they do not is because it does not...

Hi everyone,

I noticed that some people asking questions about how well the
voice-recognition works and thought I'd share my two cents. I happen
to really enjoy using the voice-recognition software, what let's
rephrase that. In comparison to having to use notetakers and writers
this is a much better choice.

I use Dragon NaturallySpeaking preferred. I've been using this
program for over 10 years and have watched enormous transformation
over the past years. a big factor for successful recognition is a
computer with a high-speed processor, significant amount of RAM, a
good microphone, and learning how to speak in a coherent matter. the
more words you can string together at a time to more successful your
recognition will be. For example, my last three sentences had no
errors but my first sentence had several. Learning how to speak for
dictation was not easy as I'm a Long Island girl. Anyone who knows
people from Long Islandbe rather fast. This does not work well with
dictation as when speaking fastwords or syllables tend to be garbled.
I notice now almost all my errors argues her errors not the program.
I must say this is only true for the 7.3 version which is at least
three or four times faster than the seven .oh version (I left this
mistake in place to show you a error which was technically a dragon
error but was contributed by myself since I paused in the midst of the
sentence.)

for me the hardest technique to try to acquire, actually I am still
trying, is the ability to say a full sentence at one time. This is
not something that is easy to do in my areana. I'm finishing my law
degree and many of the sentences we write have a great deal of meat
in each portion. Being able to say the sentence in his attire day
before dictation is almost impossible. I focused more on trying to
say 10 or so words at once. When following this philosophy as well as
speaking and enunciating each word the recognition and speed is quite
impressive. If reading out of a textbook I can easily get 70 words or
more per minute.

Also, something I've learned to increase my speed is to train each
word in the correction menu. In Dragon when using the "correct that"
command a menu box appears that allows you to correct the word. One
option also allows you to train the new word. I have found doing so
significantly decreases in this recognition of this word in the
future. I've only had on several occasions where the word and error
reappeared more than once. It was not the fault of the software but
that of my underlying accent. (I even used a command for this.) I
don't have a very strong accent but some words are affected by it and
I must train dragon with my different pronunciation.

I'm using a Pentium 4 with a CPU of a 2.4mhz and 512 ram. With this
system the software program runs rather quickly and can even play
MP3s, iTunes, and stream music while dictating. Do note in order to
do this you must use a USB microphone, I use a digital signal
processing microphone from planatronics.

I'm really excited about the prospects of the 8.0 version. For those
readers interested, this is the preferred version, the middle level
program.

I'll be really happy when the voice dictation can work with Firefox
like it works with Internet Explorer. I truly disliked the latter
program and will be a very happy camper when Dragon can asked us all
of the links via dictation as a canned in IE. For those of you
unfamiliar with this, win in IE you can use Dragon to select links by
simply saying "click and the first word of the link". If more than one
exists, number tags will appear by each and the user chooses which
link they desire.

Please note in order to use this program successfully you truly do
need a good processor with a strong CPU speed otherwise a processing
speed will become quite frustrating. If you don't have the latter
actually suggest waiting until you have the necessary hardware. As a
person who was forced unto the use of Dragon because of disabilities I
didn't have such a choice of. I can't describe the levels of
frustration I encountered. To be honest its quite amazing that a
computer never actually went through my bedroom window.

As an fyi, I owe my undgrad, graduate and soon-to-be law degree to the
use of Dragon. I actually conduct all law school exams with Dragon and
use it for writing all papers and homework assignments. Without Dragon
would not have been able to succeed in Law school since my disability
significantly limits any writing or typing abilities.

I noticed someone mentioned Via Voice. I tried it awhile back and
truly hated the program. However, this was sometime ago so he no
longer familiar with its engineering. The methods for corrections
were more tedious and time-consuming than with Dragon. To be honest
almost all my experiences with Dragon. Especially during the latter
years as each upgrade has brought significantly increased accuracy and
speed. For example must last paragraph I only had one error and it
was mostly my fault as the word was slurred.

Hope this helps!

Sorry to write such a long letter, I promise to take a step off of the
soapbox. Take care!

On 21 Nov 2004 17:30:17 -0800, (Deborah) wrote:

>Hi everyone,
>
>I noticed that some people asking questions about how well the
>voice-recognition works and thought I'd share my two cents. I happen
>to really enjoy using the voice-recognition software, what let's
>rephrase that. In comparison to having to use notetakers and writers
>this is a much better choice.
>
>I use Dragon NaturallySpeaking preferred. I've been using this
>program for over 10 years and have watched enormous transformation
>over the past years. a big factor for successful recognition is a
>computer with a high-speed processor, significant amount of RAM, a
>good microphone, and learning how to speak in a coherent matter. the
>more words you can string together at a time to more successful your
>recognition will be. For example, my last three sentences had no
>errors but my first sentence had several. Learning how to speak for
>dictation was not easy as I'm a Long Island girl. Anyone who knows
>people from Long Islandbe rather fast. This does not work well with
>dictation as when speaking fastwords or syllables tend to be garbled.
>I notice now almost all my errors argues her errors not the program.
>I must say this is only true for the 7.3 version which is at least
>three or four times faster than the seven .oh version (I left this
>mistake in place to show you a error which was technically a dragon
>error but was contributed by myself since I paused in the midst of the
>sentence.)
>
>for me the hardest technique to try to acquire, actually I am still
>trying, is the ability to say a full sentence at one time. This is
>not something that is easy to do in my areana. I'm finishing my law
>degree and many of the sentences we write have a great deal of meat
>in each portion. Being able to say the sentence in his attire day
>before dictation is almost impossible. I focused more on trying to
>say 10 or so words at once. When following this philosophy as well as
>speaking and enunciating each word the recognition and speed is quite
>impressive. If reading out of a textbook I can easily get 70 words or
>more per minute.
>
>Also, something I've learned to increase my speed is to train each
>word in the correction menu. In Dragon when using the "correct that"
>command a menu box appears that allows you to correct the word. One
>option also allows you to train the new word. I have found doing so
>significantly decreases in this recognition of this word in the
>future. I've only had on several occasions where the word and error
>reappeared more than once. It was not the fault of the software but
>that of my underlying accent. (I even used a command for this.) I
>don't have a very strong accent but some words are affected by it and
>I must train dragon with my different pronunciation.
>
>I'm using a Pentium 4 with a CPU of a 2.4mhz and 512 ram. With this
>system the software program runs rather quickly and can even play
>MP3s, iTunes, and stream music while dictating. Do note in order to
>do this you must use a USB microphone, I use a digital signal
>processing microphone from planatronics.
>
>I'm really excited about the prospects of the 8.0 version. For those
>readers interested, this is the preferred version, the middle level
>program.
>
>I'll be really happy when the voice dictation can work with Firefox
>like it works with Internet Explorer. I truly disliked the latter
>program and will be a very happy camper when Dragon can asked us all
>of the links via dictation as a canned in IE. For those of you
>unfamiliar with this, win in IE you can use Dragon to select links by
>simply saying "click and the first word of the link". If more than one
>exists, number tags will appear by each and the user chooses which
>link they desire.
>
>Please note in order to use this program successfully you truly do
>need a good processor with a strong CPU speed otherwise a processing
>speed will become quite frustrating. If you don't have the latter
>actually suggest waiting until you have the necessary hardware. As a
>person who was forced unto the use of Dragon because of disabilities I
>didn't have such a choice of. I can't describe the levels of
>frustration I encountered. To be honest its quite amazing that a
>computer never actually went through my bedroom window.
>
>As an fyi, I owe my undgrad, graduate and soon-to-be law degree to the
>use of Dragon. I actually conduct all law school exams with Dragon and
>use it for writing all papers and homework assignments. Without Dragon
>would not have been able to succeed in Law school since my disability
>significantly limits any writing or typing abilities.
>
>I noticed someone mentioned Via Voice. I tried it awhile back and
>truly hated the program. However, this was sometime ago so he no
>longer familiar with its engineering. The methods for corrections
>were more tedious and time-consuming than with Dragon. To be honest
>almost all my experiences with Dragon. Especially during the latter
>years as each upgrade has brought significantly increased accuracy and
>speed. For example must last paragraph I only had one error and it
>was mostly my fault as the word was slurred.
>
>Hope this helps!
>
>Sorry to write such a long letter, I promise to take a step off of the
>soapbox. Take care!

I appreciate your take on this but I too have been using the program
since its inception, also with high powered equipment, so we are
together on that. Where we differ is that I have been practicing law
for 25 years and it would be a godsend if this program delivered as
represented but it just does not. Perhaps you have gotten it to
perform miracles but I have been unable to. I will tell you that once
you begin practicing law, your reputation will rise or fall on the
quality of your output. One or 2 misinterpreted words that are not
caught can create a bad impression so just make sure you proofread
every word after you have dictated. I have always found that it took
more time to correct the stupid mistakes e.g. "to" instead of "two"
than it was worth. I still think we've got 5 more years to go before
it's performance becomes acceptable...