speech recognition & e-mail messages probs !!!

I am getting to grips with the Windows Vista "speech recognition" program.
Being a quadriplegic, I find it a Godsend, however, there are still some
things about the program that frustrate me and an answer to this question
would be greatly appreciated.

When I am dictating an e-mail, I seem to lose all of the punctuation and
direction commands and the speech recognition just keeps on typing what I am
saying. For example; if I say comma, or space, or backspace, it just types
the written word just like I have above and it just keeps on typing what I am
saying. Even in a message like this, it will start perfect and "text to
speech" will be working fine and then all of a sudden, "text to speech"
disappears and an alternatives panel comes up for me to choose which text I
want, which is not good for me. Can any one resolve this for me or tell me if
I am doing something wrong?

Another question: is the "Dragon Naturally Speaking" a superior program to
the "Windows Vista Speech Recognition" program and these problems I mention
above, do they or do they not arise in the Dragon program?

An answer to these questions would help me immensely.
--
andy t

2nd question, Dragon naturally speaking is better than the default vista one. Accuracy is a
little bit better, but the functionality is way better.

In ur first question, the only thing I can think is u are saying what you want to type, and
saying something that speech interprets as a command. say "what can I say" and check that
nothing ur saying is interfering with the command menus.

also might seem obvious, don't have TV or radio going lol



"andy t" <andyt@xxxxxx> wrote in message
news:96EC93F8-3420-4B65-89EC-460533282B6A@xxxxxx

--
If you don't stand behind ur soldiers, feel free to stand infront

Hi Andy,

My name is Grant, and I am a relatively long time user of Dragon NaturallySpeaking. I'm not handicapped but I use it just because of its convenience. The general answer to your question is that, (in my opinion) DNS is infinitely better than Vista speech recognition at this point in time. The reason is Dragon NaturallySpeaking came into being in the late 1990s and has been continuously improved through ten versions (Ver.10 was introduced this month). Microsoft introduced their speech recognition program about five years ago in Office 2003. Many people don't know that the Office 2003 version even exists because Microsoft never publicized it. Some people feel the 2007 version in Vista does not work as well as the older version. Actually, the recognition accuracy is equal to DNS, but the many functions are lacking because the necessary user features have not been developed sufficiently. At the rate things are going I believe it will take at least three years before Microsoft speech recognition is competitive with DNS.

What I suggest is that you join the "KnowBrainer" forum. The URL is: KnowBrainer Forums - Forums

In my opinion the KnowBrainer forum is the best source of information and help anywhere on the Internet for Dragon NaturallySpeaking and also for speech recognition in general. If you click on that page you will see that the forum has several sections. The second from the top is Naturally Speaking, for DNS. The seventh from the top is for Microsoft Vista Speech. It's moderated by Mr. Rob Chambers who actually works for Microsoft on the Vista speech project. Any questions you have about Vista speech recognition can be posted there.

The first section on the forum is devoted to "KnowBrainer" software, a program that functions with Dragon NaturallySpeaking. It consists of a large quantity of "macros", i.e. voice operated commands that make hands-free operation of your computer possible. The KnowBrainer Company is managed by Mr. Lunis Orcutt, who also moderates the KnowBrainer forum. (They also sell all versions of Dragon NaturallySpeaking and a large assortment of microphones suitable for speech recognition.) I think the retail price of KnowBrainer software is around $180 but Lunis provides free copies to handicapped people.

I'm dictating this with Dragon NaturallySpeaking Preferred, version 9.5. Although version 10 is now on the market I will not purchase it for at least three months because I have found that there are always a few bugs in the new versions :-). If you're interested in buying DNS version 9 there are some good deals right now. I see one on Amazon.com with a $50 rebate. If you buy Version 9 Preferred you can upgrade to Version 9.5 with a free download from Nuance, the maker. Information about downloading the 9.5 upgrade is available on the KnowBrainer forum. I would suggest going this route and in a few months you can upgrade to version 10. That's what I'm going to do anyway. Actually, version 9.5 works so darn well I'm not sure I want to upgrade to version 10 :-). I'm getting almost 100% one accuracy right now.

You have any questions about microphones or what kind of computer works well for DNS I'll be happy to give you my thoughts. The proper microphone is one of the most important things to achieve high recognition accuracy.

Regards,
Grant

In article <0f61035990f578c875192242e3f08b71@xxxxxx-gateway.com>,
Grantx <guest@xxxxxx-email.com> wrote:

Grant, that was a very good post to Andy.

I second all your suggestions.

Mark-


(Andy, do not read the following optional comments
under any circumstances, as they will fry
your brain. Do not say I did not warn you)

--
Will add a few observations of my own.


Warning, this is going to be long, so stop
reading now, get some snacks, find a very
comfortable chair, turn on the "Pink Panther"
on television for background. ;-)
******************************************

Newbies to speech recognition have it rough,
us old timers sometimes forget the hassle
we went through getting our skill level
up to par.

Unfortunately, there are many booby traps
associated with speech recognition, no matter
which speech app one chooses.

....and believe me, there are great practical
differences between the available speech apps.

Some really stink, from a "practical" standpoint.

For example, for Mac users, recently a brand new
speech app was created way back in April, the name
is "MacSpeech Dictate" version 1.0

All sorts of fanfare, great press releases, most
everyone had great things to say about it.

One glaring omission, there was no way for a user
to "train" the program to properly recognize a word
that the program consistently got wrong.

That is a death sentence for any speech app.

To this day, they have not added correction features.

Mac users no longer talk about MacSpeech Dictate,
they are ignoring that app, like it was never created.


Now for a change of pace, some good news
---------------------------------------

Recently, using a very high priced version of Dragon,
along with many high priced hardware attachments,
and my skill level developed over many years,
I dictated the following speech snippet
to the Dragon app.



X-Linked agammaglobulinemia (Bruton's agammaglobulinemia)
is a hereditary immunodeficiency disorder due to an
abnormality in the X chromosome and resulting in few or
no B lymphocytes and very low levels of antibodies.


The above is the raw text output, there is one mistake.

I did not have to spell anything. For example,
Dragon recognized "Bruton's agammaglobulinemia"
as being a standard malady.

Dragon made one tiny mistake that I corrected
by voice. The mistake was:

"resulting in few are no"



That had to be corrected to:

"resulted in few or no"


Dragon did an excellent job of recognizing all that
medical gibberish, because I used the "full" medical
version of Dragon.

Don't even ask me how much it cost, as I am still
in shock from paying that price several months ago.




The medical fraternity in the U.S. spends ten billion
dollars annually converting speech into text.

It obviously pays off, because they still employ it.



Now a few words about speed of operation
----------------------------------------

Most people speak in the range of 100 wpm to 200 wpm,
depending on the speaker and the complexity of the
words they are saying.

"Speed Talkers" presently can speak a bit faster
than 600 wpm.

For a new speed record to be awarded, they are given
the choice of reading a passage from the bible, or a
passage from Shakespeare, those passages picked
by the judges of course.

They have to read for a 60 second interval.

The panel of judges slows down their speech by means of
electronic devices that do not distort their speech.

PROVIDED no syllable of any word is mis-pronounced,
a new worlds record will be awarded.



Anyhow, in testing Dragon on my Mac, I managed to speak
at the rate of 336 wpm for 60 seconds.

(my tongue will never be the same again, think I broke it)

All 336 words were flawless, no mistakes, so I am happy
with the speed performance of Dragon on my Mac.



Back to speech app drawbacks
----------------------------

No matter how much you pay for a speech app, they still
make mistakes.

For example, no speech app that I am aware of can
properly handle homonyms such as: to too two


All present speech apps are afflicted with all sorts of
problems, which you do not discover until you actually
use the speech app during everyday use.

toappreciatetheproblemhereiswhatthespeechapphears

(I did that previous line deliberately, to show you
what any speech program has to contend with,
it is surprising they work as well as they do)

The same line with proper spacing and punctuation:
"To appreciate the problem,
here is what the speech app hears."



Try not to be discouraged by the deluge of stuff
associated with speech-to-text technology
----------------------------------------

Sure, it is tough, but no pain, no gain.

Pay your dues like we had to, learning, experimenting,
possibly wasting money on stuff that does not work.

(I have a top rated $400 Sennheiser microphone that
is gathering dust because it did not work for me)


What did I do? I turned right around and bought a
slightly different model $400 Sennheiser, which is
a prized possession; it picks up my soft voice at a
distance of 50 inches with amazing accuracy.

....and I am on welfare, your taxes are paying for all
of my gear, thank you.

BTW, some of my $20 microphones, the ones _included_
with Dragon, are every bit as good as that $400 mic,
as far as accuracy of text is concerned.

Good luck Andy

....now go back and re-read Grant's post, there is a lot
of good advice there.

Like Grant suggested, I would recommend starting out
with Dragon Preferred to keep the cost down until you
get your feet wet with this extremely fussy technology.

I have a copy of it myself, along with its included
microphone, it is every bit as accurate as my $1,200
version 9.5 of medical Dragon, when it comes to
everyday speech.

With effort, that $200 version of Preferred can be
made to recognize medical words such as:

pneumonoultramicroscopicsilicovolcanoconiosis

....a malady caused by breathing the fine silica dust
from an actively erupting volcano.


I myself could never again be happy with the Preferred
version, because it is missing too many useful features
for my taste.

Above all, avoid the $100 "Standard" version like
the plague, it is missing way too many useful features,
such as the ability to use the original speech audio.

Even the Preferred version will "lose the audio" if you
shut down the Dragon app, which means you have to do all
your text correction _before_ closing the Dragon app.

Sleep mode is okay, but anything that exits Dragon
will lose the audio, permanently.

That audio is _very_ _important_.

The $900 "Pro" version can permanently save the audio.

Rule of thumb: Correcting the text will take as long as
your original dictation. If you dictate for an hour,
it will take an additional hour to correct mistakes.

Chain stores like Office Depot are a good place to buy
the Preferred version.

Just be hesitant about apparent bargains online, for
fear they might be pirated versions.

Hi Andy,

This is addressed to you, regarding Dragon NaturallySpeaking (DNS) pros and cons.

Speech recognition is a very complex technology and it's not perfect by any means. But it has improved tremendously in the last 10 years and is now very useful for almost anyone willing to take a little time to learn how it works. For native English-speaking voices the "training" only takes a few minutes You do not need to buy the expensive medical version (or other "professional" versions) to get good results. The "preferred" version is the one that's used by most people for general dictation. This may sound strange but complex medical terms are more easily recognized than some words used in general vocabulary because medical versions of DNS are designed specifically to recognize them. So don't be put off by this, you don't need the medical version of DNS unless you are a doctor transcribing reports.

It's not necessary to buy the most expensive Sennheiser microphone. There are several microphones that sell for $100 or less that work fine. I'm using a headset microphone sold by RadioShack for about $20 (model 33-3012) and getting excellent recognition (usually over 98%). However for handicapped persons other types of microphones might be necessary, such as those using a microphone boom which can be positioned to operate hands-free. Also there are several wireless microphones that can be used. Possibly you're already familiar with this but if you have any questions I suggest posting a query on the KnowBrainer forum. There are many knowledgeable people participating on that forum and you will get some excellent advice.

The microphone supplied with the DNS software is a rather low quality model to keep the cost down, and does not give satisfactory results.

DNS does make mistakes with words that sound the same, such as to, too, and two. Also there and their. But it's easy to edit these words when mistakes occur.

Here's a tip: When correcting a mistake it's best to select a phrase of three or four words including the error when making the correction. If you select one word out of a paragraph DNS does not have enough data to decide which word is correct, but if you select the word in a phrase of other words you have a better chance because DNS works by associating the wrong word with nearby words to decide the most likely possibility.

Re: Vista speech recognition (VSR). The advantage of VSR is it has been designed to control many functions of Windows and other Microsoft programs with built-in voice macros. Therefore many people find it convenient to use both VSR and DNS. You can have both programs running simultaneously and switch between them. It's possible that Vista speech recognition will be your best tool in the long run -- once they sort out the other features that are lacking (the editing features DNS has). If you monitor the KnowBrainer forum section on VSR moderated by Rob Chambers, you will be able to keep up with the latest developments.

I agree you have to be careful when buying DNS from unknown vendors because there are some illegal copies being sold. I noticed that Lunis Orcutt has still a few copies of DNS Preferred ver. 9, (full version not upgrade) priced at $99 listed on his website.

Grant

Thank you very much Grant in replying to me in so much detail !

I most certainly will visit that forum you have suggested. I have some
follow-up question's and information four you.

I am quadriplegic, I had a car accident and unfortunately I broke my neck.
So you can see how important these programmes are for me.

Firstly, I should point out that I live in the UK, so do you think that
Lunis would
stretch to giving out a free copy over here?

Secondly, I have a "Dell" laptop and it has an integral Mic which is extremely
sensitive, and this works perfectly with the Pre installed
"Windows Vista
Home Premium" which came with my laptop. So if I installed a
DNS
program on my laptop, surely the integral Mic that I have
would be
sufficient would it not?

Finally, Yes I have been told that DNS 9.5 Preferred is one of the best
around, but
what about DNS Professional? Is that even better than the 9.5
version?

I find Vista's SR program a great improvement on all versions from four
years ago when I was still in hospital and it gradually learns from your
mistakes and improves itself the more you use it. The only thing that I think
needs improving is the voice commands. After reading your report, you
recommend that you have both "Vista's SR" & DNS installed so you can choose
and interact with which one you want, that sounds very cool and interesting.
Thanks again for all the advice Grant, it is a great help.
--
andy t


"Grantx" wrote:

Hi Andy,

Yes I'm pretty certain you are eligible for free copy of KnowBrainer according to what I understand from the forum. Actually it would be sent to you by e-mail so there's no shipping involved, just a program that you can download on your computer.

This is only a brief message because the hour is late and I have to get up early tomorrow :-). But the quick answer to your question about the microphone is you need to use an external microphone not the one on the laptop if you're going to use DNS. There are a dozen good reasons. This would be a good question for you to post on the forum.

Also, you do not need to buy the expensive professional version of DNS; the Preferred version has the same recognition engine as the professional version and works just as well for most purposes. The difference between the preferred and the professional version is the professional version has some programming capability, i.e. macro programming. But the KnowBrainer program is intended for macro programming; better than that it has over 10,000 preprogrammed macros (according to Lunis) that you can select from, so you don't have to learn how to write macro programs. That sounds like a lot of macros but you only need a few of them and you gradually learn which ones work best for your purposes.

When you join the forum and you are listed in the membership section it will be possible for me to send you a private e-mail and we can exchange e-mail addresses if you like. I will be happy to answer many of your questions on private e-mails and you can also get a lot of information from the forum. There are people on there that know much more about the subject than I do.

Particularly your questions about the trouble with Vista speech recognition. It sounds like there's something causing it to switch from command mode to dictation mode so that it types out the command words instead of acting on them. I have no idea what this is because I'm not an expert on Vista speech, I don't use it. But it very well could be something to do with your audio setup where the internal microphone is misinterpreting some of the sounds.

I notice that the URL I sent did not come through properly so I suggest you make a Google search for KnowBrainer and you'll find the address of the forum.

There is also an excellent source of information right in the UK from a certified Nuance vendor. He is a friend of Lunis and makes frequent postings on the forum.

I'll look for you as soon as you join the KnowBrainer forum and send you a private e-mail with my e-mail address.

Grant


PS. Please inform me of the name you use when you join the forum. My name on KnowBrainer is the same as I use here, Grantx.

Thanks for all the valuable advice Mark,

I didn't realise how complex it was. You answered a question that has been
puzzling me for some time. I did read that it was essential to use speech
recognition using a suitable microphone. As mine came with an integral Mic,
I just assumed that any speech recognition program would be fine with that.
I thought by having an external Mic as well, that they would conflict with
each other, but you have explained that this is not the case and that it is
essential to have one with the DNS program. I haven't tried your link yet,
but could you actually give me a make model number of a decent one? What
one have you got?

Having said all that, this Vista program, I find not to be all that bad, and
it works great with its integral Mic.

When reading Grant's earlier post, he said that he used both the Vista and
the DNS Programmes both for different reasons. I would be interested in
understanding how that works. I have not found time to visit the forum
"knowbrainer" yet, the one that Grant recommended, but I will do so maybe
tomorrow. In fact, Grant said he used DNS all the time, so if you read this
Grant, could you tell me again how you interact with the two programmes?
Perhaps you can tell me what Mic you use?

Do you visit that forum Mark? Do you use "Speech Recognition" frequently?

I will probably find all the answers that I require in the knowbrainer
forum. I know it's best to gather all the right criteria first, before
purchasing anything.

As I am disabled, Grant said that I would be entitled to a free DNS software
program, I cannot remember which version it was, I will have to go back to
he's post to find out which one. He said I could download it from an e-mail
that he would send to me. That sounds really great but the only thing that I
am worried about, is opening a virus up on my PC. That is no disrespect to
Grant, but it has happened to me once before. Not that there is anything
worth looking at on my PC.

I will try the forums that he mentions though, and will get to know him
first. I fully expect that he will reply to this post, it is nothing
personal and I think he will realise that.

--
andy t


"Mark Conrad" wrote:

In article <56AB0437-5AEF-4560-8CE2-ACCC3D410D7E@xxxxxx>,
andy t <andyt@xxxxxx> wrote:
Yeah, far from plug-and-play, many "gotchas" that can
trip you up, so will try to keep things simple so as
not to overwhelm you.


I for one wish my computers built-in microphone was
better, if for no other reason than I hate to wear
a headset microphone.

No, the Windows OS forces you to select the TYPE of mic
from a list, all you have to remember is to NOT select
"Internal Microphone", which is the default.

In my case, I select "USB Microphone" on that list,
being careful to avoid a similar selection which
reads "USB Array Microphone".



When you actually try the mic that is included with
Dragon, we will walk you through it, because there are
other settings in Windows itself that have
to be adjusted. Ain't complexity wonderful ;-)

You can bypass some of the mic complexity by getting
that microphone USB "translator" before you even buy
Dragon, that way you can just plug the Dragon headset
into one end of the translator.

An attached USB cable from the other end of the translator
plugs directly into the USB connector on your computer.

How bypass the complexity? Because otherwise, to get
your Dragon headset to even work with your PC, you would
have to set the Windows OS to entirely different settings,
such as "Line In Microphone" instead of "USB Microphone",
plus adjust your volume settings inside Windows OS to
entirely different values.

Best to plan ahead to avoid such unnecessary complexity.


About the translator/sound-pod/ext sound card/whatever...
-------------------------------------------------------

I gave you the first link I stumbled across.<g>

Ya pays ur money, ya takes ur chances with the people
who sell this stuff. Can't say as I really trust any of
these companies, because they all make a big case that
you need to buy a lot of their junk to do decent speech
recognition, when such is not the case.

Hmm, lemme see, what do I consider important for this
particular device.

Okay, get one that advertises "duplex operation", which
merely means that _everything_ goes through the USB
connection, coming and going.

In other words, your voice goes _into_ your computer,
and audio playback of text comes _out_ of your computer
all via the USB connector; no need to string a separate
wire to get audio playback of text-being-corrected by you.

The fewer wires strung around your PC, the better.

Actually, now that I take a closer look at my first
recommendation, (the web link I posted) - I do not
like that one. It has a "mute" button on it that
will not do you any good, because it is unlikely
that you can press that button.

Also, the attached USB cable is only a few inches long,
while we both could use a longer USB cable, say 3 feet
long, so we would not get our necks hung on too short
a cable from the headset itself.

Let me check the two different ones that I presently
own, and get back to you with an actual website where
you can buy something that will work for you.

Before I leave to check further on this device, let me
emphasis again _why_ I consider it important to use
such a device, considering the fact that it will set
you back about $50 U.S.

Output of most good microphones is very weak. If you
run that feeble signal into an electrically "noisy"
computer, other interfering signals "jump onto" the
wires carrying that weak signal, really messing up
the weak signal that is carrying your speech.

I consider it criminal that Dragon does not include
this separate device as part of their package, instead
relying on the user to have enough sense to buy
it separately.

This electrical interference is especially pronounced
if you are using a laptop, where everything is jammed
together. Nuff said.

Mark-

Hi Andy and Mark,

I agree that if you're going to use an external microphone with a laptop computer you should use a "USB sound pod" which is actually an external sound card that you connect via a USB port. The basic reason for this is that the sound system of most laptops tends to have a relatively high noise level because it picks up electrical interference from other parts of the motherboard, and this interferes with recognition accuracy. A USB sound pod is isolated from the internal noise generated by the laptop computer and consequently obtains higher recognition accuracy.

If you're using a desktop computer it is possible to use an internal sound card particularly those made by SoundBlaster and get just as good results as you will with an external USB sound card. But I believe we're talking about laptops in this discussion so I definitely agree with the idea using the USB sound pod.

I have purchased several USB sound pods, and I strongly recommend the Andrea "PureAudio USB-SA USB Pod", the latest USB sound pod product of Andrea Electronics Corporation. This is a duplex soundcard, which accommodates both your microphone and headphone or speaker. I'm using it on Windows XP and as soon as you plug it in Windows recognizes it and automatically switches all functions from the internal sound card to the external USB-SA sound pod. The sound pod has the lowest noise level of any I've tried and gave a noticeable increase in recognition accuracy, which is why I recommend it.

The model USB-SA looks like a thumb drive, or USB drive, and has no cable attachment. It plugs directly into a USB port. However I've found it very convenient to use a USB extension cable and plug the sound pod into the extension cable, which gives me extra length and prevents strain on the USB port if you plug your microphone or headphone directly into the sound pod when it's sticking out of the side of your laptop.

As far as vendors are concerned, Mark is right; there are numerous vendors advertising microphones and other equipment for speech recognition, and not all of them give you the best advice. That's why I suggest purchasing microphones and sound pods from KnowBrainer Company. I find they are the best for technical support and they also have a very good return policy if you're not happy with your product. (I have no connection with KnowBrainer; I only recommend them because I've used them several times with good results). One problem for Andy might be fact that he lives in UK, and shipping from the USA is expensive. There is a vendor in UK that you could check for price and delivery to compare with KnowBrainer. His name is Graham Hendry; his company is ITSpeaking.com, located in Birmingham.

Andy, it's great that you are achieving such excellent results with your internal microphone and Vista speech recognition. I imagine you have your laptop computer located directly in front of you so there's not much distance between you and the microphone and also you are in a very quiet room. One problem with an internal microphone is they are usually quite sensitive and tend to pick up ambient noise in the room such as radio or TV sets or other people talking. This makes it very difficult for the speech recognition program to function correctly and accuracy will drop. But if you have a low noise environment (quiet room) and you are close to the microphone it could work OK.

The advantage of external microphones made for picking up speech (as compared to picking up music from an orchestra) is they have a directional pattern so they pick the sound up mostly from the person speaking directly in front of the microphone and sounds from the sides or the back are not picked up as well. Many also have a "noise canceling" feature, which involves a complicated acoustic design, making them less susceptible to disturbance from external noises.

Mark, as to the makers of speech recognition programs supplying a proper microphone with the product, that's always been a weak point in their offering. But the problem they have is the cost of a good microphone (and possibly a USB sound pod) would equal or even exceed the cost of making the software and would likely double the selling price. In other words instead of $200 for DNS Preferred they might have to charge $400! (Remember there are several markups involved before product hits a retail shelf). However they have done something about this. The latest DNS version 10 Preferred is being offered with an optional Plantronics Corp. wireless microphone - at a higher price. This is the first time they've ever offered a good quality microphone; a few users that I've read comments from on the forum say they're very pleased with it. The Plantronics microphone clips on your ear and is battery operated, that's about all I know so far.

Now to Andy: I mentioned the model number of the headset microphone I use (made by RadioShack) but I'm not sure if you find it in the UK. However you may need a different type of microphone; I would hesitate to recommend a particular model without more information. That's why I suggested that you join the KnowBrainer forum. At last count I believe there are around 3500 members and those include people with every known type of handicap. If you were to go on that forum and post a question asking what type of microphone would be best for you (be as specific as you can about how you are using your computer etc.) you will get some very good answers about what type of microphone would be satisfactory. I expect there's probably more than one choice. Therefore I suggest you do some research on the forum and take a little time before making a final decision. Lunis sells almost every type of microphone used for speech recognition, so you could probably obtain what you need from KnowBrainer.

Grant