Discussion:
PC Speech Recognition App's on a Mac, Pros and Cons
(too old to reply)
Mark Conrad
2007-04-28 07:56:27 UTC
Permalink
Here is the unvarnished truth of what us Mac users can expect from PC
Speech Recognition app's now that we can run Windows XP Pro and Vista
Ultimate directly on our Macs, thanks to Parallels and BootCamp.

The raw uncorrected text is below, which was spoken in at 100wpm.

(138 words which took 83 seconds for me to dictate)

Start of raw uncorrected text -
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Semantic technologies are based on models that explicitly encode the
meaning of information to avoid ambiguity and support automated
reasoning. These models are usually called "ontologies". Automated
reasoning is supported to check consistency and to infer new
information. Semantic technologies provide a variety of uses and
benefits, including information integration, better search, sharing and
reuse, improved flexibility and reliability and cheaper maintenance. A
major part of this tutorial is devoted to clarifying the relationships
between the many different modeling formalisms that are the basis for
semantic technologies. These include: controlled vocabularies, the
sorry, taxonomies, ontologies, object oriented models and database
schema. We will critically evaluate the true, yet misleading claim that
ontologies have formal semantics that allow for unambiguous
representation of information. We close by considering what you can do
to get started on your own semantic technology project.

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
End of raw text -


Four mistakes in the raw text, from the beginning of example
***********************************************************

1) technologies s/b Technologies (on first line)
2) sharing and reuse s/b sharing & reuse
3) flexibility and reliability s/b flexibility & reliability
4) the sorry s/b thesauri


The first three mistakes were my fault, therefore should not really be
counted as mistakes. I have not used Dragon for almost a year on any
serious dictation project so I am a little rusty.

What I _should_ have said to eliminate the first three mistakes:

1a) "Cap next word technologies"
2a) "sharing ampersand reuse"
3a) "flexibility ampersand reliability"

So we have one "legitimate" mistake in 138 words, well within the
advertised accuracy of 99% claimed by Dragon.



The fourth mistake, "the sorry" instead of the correct "thesauri" is
a serious limitation of all present speech recognition applications,
even the very expensive Dragon NaturallySpeaking Pro versions 9.00 and
9.5

Namely, they can not handle homonyms properly.

The pronunciation of the word "thesauri" is "thu-sorr-eee" ...

... which Dragon mistakes as "the sorry".


There is no practical way to avoid such mistakes. In the distant
future, speech recognition app's will be able to avoid such mistakes,
however right now we have to live with these limitations.



The word "thesauri" is commonly used in technical publications as being
the plural of the word "thesaurus", as in the example:

"I have several thesauri on my bookshelf"

...used by some authors in place of the equally correct:

"I have several thesauruses on my bookshelf"





My particular setup here is an Intel-based MacBook Pro Duo, 2.33GHz, 3GB
RAM, Parallels build 3188, Windows XP Pro, running the Windows app'
Dragon NaturallySpeaking Pro version 9.00 from Nuance Communications Inc.

My input microphone is a "Sennheiser ME 65 / K6" placed approx' 12
inches off to the side of my mouth and slightly above, with its battery
powered internal amplifier turned off. The mic' connects via cable to
a tiny encapsulated "Parrott VX1" sound card, commonly called a
"soundpod".

Other side of the soundpod connects via a cable to the ordinary USB jack
on the MacBook.

BTW, the Sennheiser mic' worked with surprising accuracy at 60 inches
from my mouth.

I don't have the foggiest idea where the mic' is getting its power from,
being the battery switch on the mic' is turned off. I suspect the mic'
is getting power from the USB jack of the MacBook.

One reason for using the Sennheiser mic' is to free myself from the
headset that comes included with Dragon Pro.

The included headset is quite good, assuming you buy the external
Parrott VX1 soundpod for about $20, needed to get away from the
electrical noise picked up by the Macs regular internal soundcard.

Secondary reason for the Sennheiser mic' is that it allows me to speak
in a very soft voice, which is important in the hospital where I do
volunteer work.

Sound output is via the ordinary MacBook speakers.

Mark-
Wes Groleau
2007-04-28 14:14:28 UTC
Permalink
Post by Mark Conrad
Start of raw uncorrected text -
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Semantic technologies are based on models that explicitly encode the
meaning of information to avoid ambiguity and support automated
reasoning. .....
...... These include: controlled vocabularies, the
sorry, taxonomies, ontologies, object oriented models and database
schema. .......
The fourth mistake, "the sorry" instead of the correct "thesauri" is
a serious limitation of all present speech recognition applications,
even the very expensive Dragon NaturallySpeaking Pro versions 9.00 and
9.5
Namely, they can not handle homonyms properly.
The pronunciation of the word "thesauri" is "thu-sorr-eee" ...
... which Dragon mistakes as "the sorry".
There is no practical way to avoid such mistakes. In the distant
future, speech recognition app's will be able to avoid such mistakes,
however right now we have to live with these limitations.
There are at least three ways to avoid such mistakes. (Let's not have
an argument over the meaning of "practical.")

1. Semantic technologies. If the Dragon's Dictionary included parts
of speech, and if those had been consulted, the program would "know"
that 'sorry' is not a noun and therefore not likely to be preceded
by 'the.' And analyzing the text for hints to meaning would make it
possible to know that "sorry" is isn't at home as the middle member
of a list of nouns of that sort.

2. "Training." Dragon, like ViaVoice, "learns" during correction.
Statistical methods store information from which can be computed
the probability for a certain word to be in the vicinity of certain
other words. This is similar to the Bayesian classification at the
core of any spam filter that actually works. ViaVoice (and I suspect
Dragon also) allows one to pre-train by scanning files of text
already composed. It also allows multiple user profiles, which
could be used to separately store the probability sets for different
authors.

3. Put "thesauri" in the Dragon Dictionary. :-)

Bayesian classification, like that done for spam, has been used to
assist in identifying the authors of unattributed text. Scan all works
known to be by Joe Blow, and you can generate a probability that the
next file is by him.

Google is currently working on using a variation of this to improve the
quality of machine translation:

http://www.theregister.co.uk/2007/03/29/google_translator/

The basic idea has been around for a while:
http://tinyurl.com/25qkon (Google Scholar)
http://tinyurl.com/yp8rlz (Google Popular)

As for whether "thesauri" or "thesauruses" is better, you'll just have
to _guess_ on my opinion. Hint: I fully agree with Silver Han's usage
rules, as stated in the middle of http://tinyurl.com/37afqf
--
Wes Groleau

In any formula, constants (especially those obtained
from handbooks) are to be treated as variables.
Mark Conrad
2007-04-28 21:02:44 UTC
Permalink
Post by Wes Groleau
There are at least three ways to avoid such mistakes. (Let's not have
an argument over the meaning of "practical.")
1. Semantic technologies. If the Dragon's Dictionary included parts
of speech, and if those had been consulted, the program would "know"
that 'sorry' is not a noun and therefore not likely to be preceded
by 'the.' And analyzing the text for hints to meaning would make it
possible to know that "sorry" is isn't at home as the middle member
of a list of nouns of that sort.
Yeah, I am surprised that approach is apparently ineffective with Dragon.

Homographs such as "bow and bow" screw up that approach a bit also.

"bow" can be a verb or a noun

...or even a phrasal verb, as in "bow out of"

...or even a phrasal noun, as in "shot across the bow"
Post by Wes Groleau
2. "Training." Dragon, like ViaVoice, "learns" during correction.
Statistical methods store information from which can be computed
the probability for a certain word to be in the vicinity of certain
other words. This is similar to the Bayesian classification at the
core of any spam filter that actually works. ViaVoice (and I suspect
Dragon also) allows one to pre-train by scanning files of text
already composed. It also allows multiple user profiles, which
could be used to separately store the probability sets for different
authors.
That works great for users who commonly use words like "thesauri" 95% of
the time, not so great if there is a 50-50 chance that they meant "the
sorry" or "the sore eye".

As regards adjacent words, I think I read somewhere that Dragon only
looks at the immediately adjacent word, in contrast to humans who can
listen to entire phrases, sentences, even paragraphs, to decipher
meaning.
Post by Wes Groleau
Scan all works known to be by Joe Blow, and you can generate
a probability that the next file is by him.
I personally do not bother to scan _any_ documents at all, because
almost all of my input audio varies so radically as regards subject
matter. i.e., there seems to be no improvement in accuracy when I scan
documents.

Scanning documents works great if a large part of the input audio is
consistent, say for example generating business emails about a specific
product line.



As for me, I would guess that 90% of my correction workload is because
Dragon can not yet adequately handle:

1) Homonyms, such as: two, 2, to, too
2) Homophones, such as: new, knew - - - need, kneed
3) ...and to a minor extent homographic phrases, mentioned earlier



As regards competition against Nuance, Inc.

Nuance also really likes to jack up the price of Dragon for minor
improvements, then take the profits and funnel them into eyewash
marketing for dumb users like me, instead of plowing the money back into
research and actual improvement of Dragon.

I wish MacSpeech would get on the ball and give Nuance some serious
competition.
Michelle Steiner
2007-04-28 14:52:04 UTC
Permalink
Post by Mark Conrad
The pronunciation of the word "thesauri" is "thu-sorr-eee" ...
I pronounce it "the sore eye" with a soft "th".
--
Support the troops: Bring them home ASAP.
Joey DoWop Dee
2007-04-28 14:56:36 UTC
Permalink
Post by Mark Conrad
The pronunciation of the word "thesauri" is "thu-sorr-eee" ...
... which Dragon mistakes as "the sorry".
I'm surprised that it can't distinguish between an unvoiced "th" as in
"thesauri" and a voiced "th", such as in "the sorry."

I wonder if the [bastardized or correct?] pronunciation of "thu-sorr-eye"
would have made it work?
--
Joey DoWop Dee
Remember: It is To Laugh
Mark Conrad
2007-04-28 21:02:42 UTC
Permalink
Post by Joey DoWop Dee
I wonder if the [bastardized or correct?] pronunciation of
"thu-sorr-eye" would have made it work?
It probably would have produced "the sore eye". ;-)
Post by Joey DoWop Dee
I'm surprised that it can't distinguish between an unvoiced "th" as in
"thesauri" and a voiced "th", such as in "the sorry."
Dragon and other speech app's have a lot of trouble with short articles
like "the".

Somewhere in the Dragon documentation, it sez that Dragon will not even
allow correction of "the", as for example with a religious user who
wants to use "thee" 100% of the time.


For one thing, most people commonly misuse both the "thu" and "thee"
forms of the word.

Supposedly, "thu" is correct usage before a vowel, as in "thu
asparagus", while "thee" is used before a consonant, as in "thee
pineapple".

YMMV ;-)

Mark-
Joey DoWop Dee
2007-04-28 21:14:42 UTC
Permalink
Post by Mark Conrad
Supposedly, "thu" is correct usage before a vowel, as in "thu
asparagus", while "thee" is used before a consonant, as in "thee
pineapple".
YMMV ;-)
MMDV! (definitely varies) as I've always done just the (make that "thee")
opposite: when the (make that "thu") first letter is a vowel, I say "thee"
and vice-versa.

And I always thot I tawked good!
--
Joey DoWop Dee
Remember: It is To Laugh
Mark Conrad
2007-04-29 03:05:16 UTC
Permalink
Post by Joey DoWop Dee
Post by Mark Conrad
Supposedly, "thu" is correct usage before a vowel, as in "thu
asparagus", while "thee" is used before a consonant, as in "thee
pineapple".
YMMV ;-)
MMDV! (definitely varies) as I've always done just the (make that "thee")
opposite: when the (make that "thu") first letter is a vowel, I say "thee"
and vice-versa.
And I always thot I tawked good!
Yes, even the definition breaks its own rules, as in "Thee expert".

I say whichever I feel like at the time, usually favoring "thee".

More cultured, y'know.


Altho' at times I revert to "da dummy", revealing my true uncouth past.

Mark-
Wes Groleau
2007-04-29 05:35:42 UTC
Permalink
Post by Mark Conrad
Post by Joey DoWop Dee
I wonder if the [bastardized or correct?] pronunciation of
"thu-sorr-eye" would have made it work?
It probably would have produced "the sore eye". ;-)
Post by Joey DoWop Dee
I'm surprised that it can't distinguish between an unvoiced "th" as in
"thesauri" and a voiced "th", such as in "the sorry."
Dragon and other speech app's have a lot of trouble with short articles
like "the".
Somewhere in the Dragon documentation, it sez that Dragon will not even
allow correction of "the", as for example with a religious user who
wants to use "thee" 100% of the time.
For one thing, most people commonly misuse both the "thu" and "thee"
forms of the word.
What "most people" say and accept is _not_ misuse.
There is no Emperor of English whom all must obey.
Post by Mark Conrad
1) Say it in a way that the listener will understand
2) Say it in a way that won't make you look like an idiot
As far as I'm concerned, that's as close to "rules"
as anybody needs to get.
--
Wes Groleau

Beware of the man who works hard to learn something, learns
it, and finds himself no wiser than before ... He is full of
murderous resentment of people who are ignorant without having
come by their ignorance the hard way.
-- Kurt Vonnegut
Mark Conrad
2007-04-29 16:08:58 UTC
Permalink
Post by Wes Groleau
Post by Mark Conrad
For one thing, most people commonly misuse both the "thu" and "thee"
forms of the word.
What "most people" say and accept is _not_ misuse.
There is no Emperor of English whom all must obey.
Ah yes, the Tower of Babble approach, say whatever one likes to promote
misunderstanding and confusion.

Mark-
Wes Groleau
2007-04-29 23:08:11 UTC
Permalink
Post by Mark Conrad
Post by Wes Groleau
What "most people" say and accept is _not_ misuse.
There is no Emperor of English whom all must obey.
Ah yes, the Tower of Babble approach, say whatever one likes to promote
misunderstanding and confusion.
As you have just illustrated, that is the _Usenet_ approach.
"What most people say and accept" is not as likely to promote
misunderstanding and confusion as pompously and pedantically
insisting on holding to the "proper" usage of our grandparents.

Another "Usenet approach" you illustrate here:
1. Quote part of someone's message and accuse them of
saying the opposite of the part you snip.
Post by Mark Conrad
Post by Wes Groleau
1) Say it in a way that the listener will understand
2) Say it in a way that won't make you look like an idiot
As far as I'm concerned, that's as close to "rules"
as anybody needs to get.
(Which is not to say that I disapprove of HONEST editing.)
--
Wes Groleau

Pat's Polemics = http://Ideas.Lang-Learn.us/barrett
Mark Conrad
2007-04-30 06:41:30 UTC
Permalink
...<snip>..."Usenet approach" you illustrate...<snip>...
I am not going to get into a pissing contest about this.

The nice thing about Usenet is that different views are tolerated by
_most_ reasonable people.

Few things are black and white; there are good reasons to standardize
pronunciation as much as is reasonably possible; there are also good
reasons to avoid rigid standardization, as you appeared to dogmatically
be insisting on in your posts.


BTW, found out I was wrong; all my life I pronounced "via" as vi-uh.

Dragon flatly refused to recognize my mispronunciation, so I tried
pronouncing it correctly as "vi", which made Dragon happy.

(apparently "vee" is "correct" also, according to the OS X dictionary.



I _finally_ got Dragon to recognize my "thesauri" about 50% of the
time, by pronouncing it as:

thu-SOAR-i


(thu as in thermal - - - i as in iodine)


Appears I have been advocating the Dragon beast a bit too much, so I am
going to crank out a separate post about the drawbacks of using Dragon,
so far as the usual unwashed Mac user is concerned.

Mark-
Michelle Steiner
2007-04-30 16:10:25 UTC
Permalink
Post by Mark Conrad
BTW, found out I was wrong; all my life I pronounced "via" as vi-uh.
Dragon flatly refused to recognize my mispronunciation, so I tried
pronouncing it correctly as "vi", which made Dragon happy.
Dictionaries say that the correct pronunciation is either "vee uh" or
"vahy-uh". Dragon is wrong.
--
Support the troops: Bring them home ASAP.
Mark Conrad
2007-04-30 17:17:15 UTC
Permalink
Post by Michelle Steiner
Post by Mark Conrad
BTW, found out I was wrong; all my life I pronounced "via" as vi-uh.
Dragon flatly refused to recognize my mispronunciation, so I tried
pronouncing it correctly as "vi", which made Dragon happy.
Dictionaries say that the correct pronunciation is either "vee uh" or
"vahy-uh". Dragon is wrong.
Wow, thanks for that.

Would you please check the dictionary in OS X, somehow I got the
impression there was only one syllable from that dictionary.

Probably I misunderstood it.

Thanks,

Mark-
Michelle Steiner
2007-04-30 17:27:18 UTC
Permalink
Post by Mark Conrad
Would you please check the dictionary in OS X, somehow I got the
impression there was only one syllable from that dictionary.
It shows too alternative two-syllable pronunciations; Ve (with a bar
over the "e")-uh (indicated by a backwards "e"), and vi(with a bar over
the "i")-uh (same symbol for "uh").
--
Support the troops: Bring them home ASAP.
Mark Conrad
2007-05-01 02:55:45 UTC
Permalink
Post by Michelle Steiner
Post by Mark Conrad
Would you please check the dictionary in OS X, somehow I got the
impression there was only one syllable from that dictionary.
It shows too alternative two-syllable pronunciations; Ve (with a bar
over the "e")-uh (indicated by a backwards "e"), and vi(with a bar over
the "i")-uh (same symbol for "uh").
Ah, gotcha.

I was confused by the "bullets" in the bold print, i.e. there were _no_
bullets at all for "via" all by itself, so I assumed it was only one
syllable.

Typing in "via" by itself to the OS X dictionary reveals that all the
other "via" words listed have bullets, indicating two or more syllables.

"via" all by itself was the only exception, i.e. no bullets, therefore
only one syllable to my warped way of thinking. ;-)

Mark-
Dave Balderstone
2007-04-30 18:45:24 UTC
Permalink
Post by Mark Conrad
Post by Michelle Steiner
Post by Mark Conrad
BTW, found out I was wrong; all my life I pronounced "via" as vi-uh.
Dragon flatly refused to recognize my mispronunciation, so I tried
pronouncing it correctly as "vi", which made Dragon happy.
Dictionaries say that the correct pronunciation is either "vee uh" or
"vahy-uh". Dragon is wrong.
Wow, thanks for that.
Would you please check the dictionary in OS X, somehow I got the
impression there was only one syllable from that dictionary.
Why do you need someone to check the dictionary in OS X?

According to your headers, you are running OS X.
--
"Let's just admit that public education is mediocre at best." -- Frank Zappa
Mike Rosenberg
2007-05-01 02:13:03 UTC
Permalink
Post by Dave Balderstone
Post by Mark Conrad
Would you please check the dictionary in OS X, somehow I got the
impression there was only one syllable from that dictionary.
Why do you need someone to check the dictionary in OS X?
According to your headers, you are running OS X.
I can't believe I'm going to admit to this, and in a public forum yet,
but I think I actually understood what he wrote. If I'm not mistaken,
he meant that he had already looked at the Dictionary in OS X and come
away believing it indicated that "via" is pronounced as if only one
syllable, so he was hoping someone else would confirm or deny that.
--
<http://designsbymike.biz/macconsultshop.shtml> Mac-themed T-shirts
<http://designsbymike.biz/musings.shtml> Humorous/muckraking T-shirts
<http://designsbymike.biz/prius.shtml> Prius shirts & bumper stickers
<http://cafepress.com/comedancing> Ballroom dance-themed shirts & gift
Wes Groleau
2007-05-01 03:21:22 UTC
Permalink
Post by Mark Conrad
Few things are black and white; there are good reasons to standardize
pronunciation as much as is reasonably possible; there are also good
reasons to avoid rigid standardization, as you appeared to dogmatically
be insisting on in your posts.
I thought Silver's rules (which I endorse) addressed both extremes.
--
Wes Groleau
Heroes, Heritage, and History
http://freepages.genealogy.rootsweb.com/~wgroleau/
Mark Conrad
2007-04-30 06:41:34 UTC
Permalink
Drawbacks of using Dragon NaturallySpeaking on an Intel-based Mac


Present versions are 9.00 for XP Pro, and 9.5 for Vista Ultimate.

Nuance Technologies Inc. is the present "owner" of Dragon.




Price, price, price, the high end Dragons are expensive.

$900 for the Pro version, with all the bells and whistles.

$1,200 for the medical and legal versions.

Let's not forget that clueless guys like me also get sucked in to buy a
lot of accessories like expensive microphones for $600.

One such expensive mic' did not work out at all for me, so I blew
another $600 on a different mic', which _did_ improve my setup.

FWIW, the $20 microphone headset that is included free with Dragon is
very good, excellent accuracy is obtained.

However...

Best accuracy is _not_ obtained with this free headset _unless_ you
buy an optional $20 external "soundpod" (soundcard) - which completely
bypasses the regular soundcard built into your Mac.

Internal soundcard inside the Mac picks up too much electrical
interference, which degrades the accuracy of the speech app'.

This (necessary) encapsulated soundcard can be bought at places like
eMicrophone, the model is "Parrott VXI"

Mic' and earpiece cables both plug into one side of soundpod, USB cable
on the other side goes directly to the USB jack on your Mac.



Even the lower priced versions of Dragon are fairly high.

Asking price for "Preferred" version is $200 from Nuance.

I have seen the Preferred version 9.00 advertised as low as $118 at
Amazon.

Asking price for the lowest "Standard" version 9.00 is $99.

Have seen it as low as $78 at some resellers.



High volume resellers like Office Depot and Best Buy usually have the
"Standard" and "Preferred" versions in stock, but never the "Pro"
version.



Back to the drawbacks -

Legitimate reasons for using speech recognition app's are few and far
between, usually merely typing in documents is much less costly and much
more "practical".

Even if you speak quickly at 160wpm for example, your average speed of
getting speech to text will be far below that, because the raw text has
to be "corrected" before it is usable.

For complex speech input, I consider myself lucky if I can average 50wpm.



I consider the "Standard" version of Dragon next to useless, not much
better than iListen as far as dictation is concerned.

(I have all 3 versions of Dragon)

Main reason I do not like the Standard version is that it does not
provide feedback of the original audio, which often screws up your
efforts to correct Dragon when Dragon makes a mistake.

The mid-range "Preferred" version _does_ provide the original speech
audio, however you have to be careful to _not_ exit Dragon, otherwise
the original audio will be lost forever.

(Pro version of Dragon hangs on to original audio)



Despite its higher price, the "Preferred" version seems to be the most
popular version, when people get to try all 3 versions ahead of time.





Last and most serious drawback of Dragon, and all other speech app's for
that matter, is that they are nowhere as "accurate" as a human is as
regards converting speech to text.

Humans can easily comprehend what a person in a crowded noisy restaurant
is saying, mentally weeding out all other interfering noises and
understanding what one person is saying, even when other people at his
table are talking at the same time.

Speech app's can not yet do this, it may be decades before they can.

Humans can "handle" people with different accents, different voice
pitches, different everything - - - speech app's generally only work
well when they are trained to one specific person's voice.

If you train Dragon to your voice, there is no guarantee that it will
work well when another person uses it.



Despite all the above severe drawbacks, there are situations where it
pays off to have a good speech recognition app' available.

Now, thanks to Parallels and BootCamp, Mac users can at last take
advantage of the best Windows speech app's, both the Windows XP Pro
versions, and the Vista Ultimate versions.

There seem to be fewer and fewer reasons for me to use my one PC, it is
lately just sitting there gathering dust.

Mark-
RW
2007-04-30 14:18:46 UTC
Permalink
Post by Mark Conrad
One such expensive mic' did not work out at all for me, so I blew
another $600 on a different mic', which _did_ improve my setup.
FWIW, the $20 microphone headset that is included free with Dragon is
very good, excellent accuracy is obtained.
Best accuracy is _not_ obtained with this free headset _unless_ you
buy an optional $20 external "soundpod" (soundcard) - which completely
bypasses the regular soundcard built into your Mac.
Internal soundcard inside the Mac picks up too much electrical
interference, which degrades the accuracy of the speech app'.
This (necessary) encapsulated soundcard can be bought at places like
eMicrophone, the model is "Parrott VXI"
What type of mics work best?

Is noise cancellation a priority?
Mark Conrad
2007-04-30 17:10:22 UTC
Permalink
Post by RW
What type of mics work best?
Short answer is to use the headset that is included free, however buy an
optional $20 USB "soundpod" to bypass the Macs regular internal
soundcard.

(a Parrott VXI USB soundpod is an excellent choice)

Mac's internal soundcard is great for general Mac usage, but
unfortunately it picks up a lot of electrical interference inside the
Mac's chassis.

I find that the free headset mic' has to be no further away than a
finger's width from the corner of my mouth. Even then one has to speak
at an "assertive" volume, else speech recognition accuracy falls way
down.




ONLY benefits of going to the $600 mic' like I use is that it allows you
to speak in a really soft voice, at distances up to about four feet,
while still obtaining much higher recognition accuracy than would be
possible with iListen, for example.

iListen is a speech recognition app' that only runs on a Mac, not on a
PC. iListen is from MacSpeech, Inc.

I commonly use a very expensive Sennheiser at 8 to 12 inches for maximum
accuracy. I pay very little attention to the exact distance, it often
creeps up to 18 inches without me noticing it.

Mic' is on a portable stand, off to the side and slightly above my mouth.

I also have one of those counter-balanced microphone booms, which I will
mount on my computer desk in the near future. They allow me lean back
in my chair, grab the mike and roughly reposition it.

Only important caution is to aim the mic' at the speakers mouth;
distance is much less important with this particular mic'.
Post by RW
Is noise cancellation a priority?
Long answer coming up, sorry about that.

Short answer is yes, noise cancellation is a priority.

The exception is when you are in a really quiet environment with no hard
walls that reflect confusing echoes. Close nearby walls are tolerable,
because the echoes from those are almost instantaneous, therefore can't
interfere appreciably with the speakers voice.

A number of good (expensive) microphones that are better suited for
music and stage entertainers might not be so great for speech
recognition.

...otherwise we could just buy good Mic's at any microphone store and
our problems would be solved.


Not that easy, one often has to resort to experimentation, or contact
Dragon tech' support and hope for a knowledgeable technician.





There are _true_ noise canceling mic's, the kind that actually pick up
the background noise, then electronically cancel out that noise.

Something like the same principle the "Bose" headsets used by airline
passengers, to cancel out the noise of the aircraft engine, while
listening to music.

I do not know how to test a mic' to determine whether it has "true"
noise canceling properties, or is merely directional.


Unfortunately, a lot of mic's claim to have noise canceling when they
actually only have _directional_ microphone pickup patterns.

Directional pickup _is_ important, in order to reject sounds that do
not come from the direction of the speakers mouth.

Noise canceling and directional mic's do improve the accuracy of speech
recognition, at least in my experience.

Even the free mic' headsets that are included with Dragon are
directional. They often have a white dot on the tiny mic' pickup that
tells the user which direction to orient the mic' to make certain it is
positioned properly to favor the user's mouth.




In theory, a highly directional shotgun microphone, the kind used by the
CIA, could pick up the voice of one person at a distance, convert that
person's voice to text, and ignore the voices of other people near the
target person.

That would be of value for their own agents, to enable speech-to-text
monitoring without needing to wear a "wire".

In summary, for speech recognition one ultimate test is to try the
microphone in question, to see for yourself whether or not it improves
your overall speech recognition accuracy.

Unfortunately, that approach is expensive.

I had a $400 Sennheiser MD 431 II mic' work so-so, while a free
microphone headset that came included with a $78 Dragon app' worked just
as well for all practical purposes.

Mark-
Wes Groleau
2007-05-01 03:30:05 UTC
Permalink
Post by Mark Conrad
Post by RW
What type of mics work best?
Short answer is to use the headset that is included free, however buy an
optional $20 USB "soundpod" to bypass the Macs regular internal
soundcard.
The Andrea mike that came with ViaVoice goes through the USB port.
Worked well if the room was silent, i.e., the furnace, water heater,
washer, and dryer idle. :-)

Unfortunately now, my dictation has to compete with the fans on
a Linux tower and a FreeBSD tower less than three feet away.
(One's our extra disk space, the other our DNS server and web proxy)
Post by Mark Conrad
There are _true_ noise canceling mic's, the kind that actually pick up
the background noise, then electronically cancel out that noise.
It helps if the initial training setup is done with the SAME noise
spectrum that will occur during dictation.
--
Wes Groleau

Armchair Activism: http://www.breakthechain.org/armchair.html
Mark Conrad
2007-05-01 16:34:03 UTC
Permalink
Post by Wes Groleau
Post by Mark Conrad
Post by RW
What type of mics work best?
Short answer is to use the headset that is included free, however buy an
optional $20 USB "soundpod" to bypass the Macs regular internal
soundcard.
The Andrea mike that came with ViaVoice goes through the USB port.
Worked well if the room was silent, i.e., the furnace, water heater,
washer, and dryer idle. :-)
Yeah, background racket screws stuff up. I like the Andrea headset.

Wish ViaVoice would have stayed active. No one has maintained or
updated ViaVoice for years now.


Nuance (Dragon) has a nasty trap in their website. They peddle ViaVoice
to Mac users, most of whom are totally clueless about ViaVoice being
stagnant for years.

Nuance is sneaky in other ways also, the headset they ship with Dragon
Preferred has the headset makers name is totally removed - - - Nuance
also lies a little when they claim that Dragon needs no training. The
app' will not allow one to dictate unless 5 minutes of training occurs,
even after you check the box that asks if you want to skip training.


For me, the minimum 5 minutes of training yielded results that were
_much_ better than iListen after hours of training.

Of course, I did the Right Thing with Dragon and trained it using ALL
the lessons, which took many hours/days of tedious training.

Not finished even yet, because I need to make certain that Dragon
understands every possible command it has, plus a slew of commands that
I will create myself. (possible with the Pro version of Dragon)
Post by Wes Groleau
Post by Mark Conrad
There are _true_ noise canceling mic's, the kind that actually pick up
the background noise, then electronically cancel out that noise.
It helps if the initial training setup is done with the SAME noise
spectrum that will occur during dictation.
Yes, but it might not work too well if the noise varies in nature during
the training, as in your case where different household devices keep
cutting in and out.

I have a noisy dehumidifier that I shut off prior to a dictation session.

Refrigerator does not seem to degrade the accuracy, because it is fairly
quiet.
RW
2007-05-01 15:13:31 UTC
Permalink
Post by Mark Conrad
Post by RW
What type of mics work best?
Short answer is to use the headset that is included free, however buy an
optional $20 USB "soundpod" to bypass the Macs regular internal
soundcard. (a Parrott VXI USB soundpod is an excellent choice)
Mac's internal soundcard is great for general Mac usage, but
unfortunately it picks up a lot of electrical interference inside the
Mac's chassis.
What about the iMic USB audio adapter? Does that bypass the Mac's
internal soundcard? And are all Mac built in soundcard's of similar
quality?

These are cheap and claim noise cancellation.
http://tinyurl.com/2nknsc
Anybody ever used one?
What would be useful would be to have a boom mic that you could place
over your head or ear so the positioning to your mouth would be
consistent as you move. Then to be able to put it on a stand when
you're sitting at your desk and not moving. Anybody seen that yet?
Actually, if a desk mounted mic was ok in picking up your voice (in a
quiet environment of course) within a foot, that would give ample
freedom for most people so they wouldn't have to wear it. I never
found any feedback using a softphone/VOIP mic/speakers setup. Maybe
its the lag of the VOIP that helps. It was very pleasant hearing the
other end through quality speakers. Sort of like talking to someone
right beside you. Which is cool when you're Skyping to someone on the
other side of the world for free.
Mark Conrad
2007-05-02 09:42:32 UTC
Permalink
Post by RW
Post by Mark Conrad
Post by RW
What type of mics work best?
Short answer is to use the headset that is included free, however buy an
optional $20 USB "soundpod" to bypass the Macs regular internal
soundcard. (a Parrott VXI USB soundpod is an excellent choice)
Mac's internal soundcard is great for general Mac usage, but
unfortunately it picks up a lot of electrical interference inside the
Mac's chassis.
What about the iMic USB audio adapter? Does that bypass the Mac's
internal soundcard?
Short answer is yes.

You hafta understand the basic purpose of any soundcard. It is to take
the incoming sound and convert it to digital ones and zeroes that the
computer CPU can handle and process.

There always has to be some "device" between the microphone and the USB
jack. This device can be called an "audio adapter", a "soundpod", or
an "external soundcard". (confusing, isn't it)


If all you have is a "ordinary" microphone without any "device", then
and only then the internal Mac's soundcard is used.

A separate pin jack connects the "ordinary" microphone directly to the
Mac's internal sound card. USB connector not used at all.
Post by RW
And are all Mac built in soundcard's of similar quality?
No, however do not confuse quality with electrical interference.

If there was NO electrical interference, the Mac's internal soundcard
would be more than adequate for speech recognition.


However there _is_ electrical interference, because the Mac's
soundcard is buried right next to one heck of a lot of electrical
circuitry, and microphone signals are notoriously weak.

The electrical interference therefore tends to interfere with the weak
microphone signal, before the soundcard gets a chance to amplify that
weak microphone signal.

This electrical noise manifests itself as a hisss, which degrades the
signal-to-noise ratio of the audio. i.e. with no audio input, you still
hear the hisss coming out of the speakers.

Normally this electrical noise does not bother the general use of a
microphone, where you are merely using the mic' to pick up music and
singing to be sent to speakers etc.

HOWEVER, speech recognition app's are another matter altogether, because
they are a lot more fussy as regards electrical interference.

Speech app's have to know the difference between the "ssss" sounds of
words, and the "ssss" that is the result of electrical hissing.

Almost impossible to tell them apart, so the speech app' makes mistakes.
Post by RW
These are cheap and claim noise cancellation.
http://tinyurl.com/2nknsc
Anybody ever used one?
Ya pays your money, ya takes ur chances.

I used to thrash around, bought mic's that "looked good", wasted as much
as $400 on one expensive Sennheiser mic'.

Finally wised up and listened to the guys who actually played with
speech recognition app's, as regards which microphones performed well.

Had much better luck after that.<g>



FWIW, the cheap headsets that come free with a lot of Dragon app's
worked well for me, _provided_ I bought the optional USB
adapter/soundpod/soundcard for an additional $20.

My particular USB soundpod is a Parrott VXI, bought from eMicrophones,
however I have used other brands of USB soundpods such as Andrea, and
they seem to work well also.

My only objection to all free headset mic's is that the mic' pickup has
to be practically brushing your lips in order to get good accuracy.

For mic's that work at reasonable distances, it has been my limited
experience that you will have to shell out big bucks.
($400-$600 depending on stands, attachments, etc.)

BEWARE: Most expensive mic's will not work at those distances.

I was going by the recommendation of another speech user when I bought
the microphone below:

I have a Sennheiser ME 65/K6 that I can use at 50 inches, and it will
still outperform the accuracy possible with iListen.
(using iListen's included headset mic')

Can't figure out why it works so well at that distance, because the
sound playback sounds like I am at the bottom of a well, because of
echoes off nearby walls in my computer den.

For best accuracy, I use it in the range of 8" to about 18", does not
appear to be fussy at all as regards exact distance.
Post by RW
What would be useful would be to have a boom mic that you could place
over your head or ear so the positioning to your mouth would be
consistent as you move. Then to be able to put it on a stand when
you're sitting at your desk and not moving. Anybody seen that yet?
You are speaking big bucks. A spring loaded boom will set you back
about $200, plus you have to figure on mounting difficulties.

Would be nice if one had deep pockets, I have to agree.

A flexible gooseneck mic' mount with a homemade plywood base is nice,
that is what I am presently using.

Mark-
RW
2007-05-02 15:29:27 UTC
Permalink
Post by Mark Conrad
Post by RW
Post by Mark Conrad
Post by RW
What type of mics work best?
Short answer is to use the headset that is included free, however buy an
optional $20 USB "soundpod" to bypass the Macs regular internal
soundcard. (a Parrott VXI USB soundpod is an excellent choice)
Mac's internal soundcard is great for general Mac usage, but
unfortunately it picks up a lot of electrical interference inside the
Mac's chassis.
What about the iMic USB audio adapter? Does that bypass the Mac's
internal soundcard?
Short answer is yes.
So is the iMic any different from any other audio to usb adaptor?
Post by Mark Conrad
Post by RW
And are all Mac built in soundcard's of similar quality?
No, however do not confuse quality with electrical interference.
If there was NO electrical interference, the Mac's internal soundcard
would be more than adequate for speech recognition.
However there _is_ electrical interference, because the Mac's
soundcard is buried right next to one heck of a lot of electrical
circuitry, and microphone signals are notoriously weak.
The electrical interference therefore tends to interfere with the weak
microphone signal, before the soundcard gets a chance to amplify that
weak microphone signal.
This electrical noise manifests itself as a hisss, which degrades the
signal-to-noise ratio of the audio. i.e. with no audio input, you still
hear the hisss coming out of the speakers.
I thought that was from the amp from my powered speakers but maybe its
from the Mac. I'll try another setup to see later. That hiss is more
important than most people realize because you're sitting so close to
the speakers, or at least most people do.
Post by Mark Conrad
FWIW, the cheap headsets that come free with a lot of Dragon app's
worked well for me, _provided_ I bought the optional USB
adapter/soundpod/soundcard for an additional $20.
Does anyone know who makes these OK headsets?
And does anyone know of a desktop mic (they usually use a 8" wand like
stand) that has decent performance? That means in a quiet
environment.
And does noise cancellation degrade audio quality in some way?
Would it better to use an ordinary, quality mic with a USB input in a
very quiet environment?
Post by Mark Conrad
My particular USB soundpod is a Parrott VXI, bought from eMicrophones,
however I have used other brands of USB soundpods such as Andrea, and
they seem to work well also.
How would that compare to an iMic?
Post by Mark Conrad
My only objection to all free headset mic's is that the mic' pickup has
to be practically brushing your lips in order to get good accuracy.
So if the mic had say 12" or so of leeway, that might be a good
compromise for most users. Its not as if we're doing calisthenics
when using it. :)
Post by Mark Conrad
Post by RW
What would be useful would be to have a boom mic that you could place
over your head or ear so the positioning to your mouth would be
consistent as you move. Then to be able to put it on a stand when
you're sitting at your desk and not moving. Anybody seen that yet?
You are speaking big bucks. A spring loaded boom will set you back
about $200, plus you have to figure on mounting difficulties.
I meant a mic that attaches to your head/ear/somehow and is always
positioned in front of your mouth no matter where you move, that could
be placed in a desk mounted holder in front of your monitor at other
times. In fact, having the mic just extend directly out from the
bottom of the monitor to your mouth would be great as it would be most
always in the right position and off the desk. I've never seen any
setup like that. Has anybody else? The monitor, speakers, mic,
keyboard and mouse should all move together, ideally. What would be
interesting is to have this stuff attach to a chair, so it moves with
you. Some interesting photos on Flickr using this idea.
Post by Mark Conrad
Would be nice if one had deep pockets, I have to agree.
A flexible gooseneck mic' mount with a homemade plywood base is nice,
that is what I am presently using.
Sounds like a good, workable solution.
Mark Conrad
2007-05-02 17:54:21 UTC
Permalink
Post by RW
Post by Mark Conrad
hear the hisss coming out of the speakers.
I thought that was from the amp from my powered speakers but maybe its
from the Mac.
Could be from the amp', I am just going by what the so-called experts
say about the interference pick up by the sound card.


You have the right attitude. Do not take anything you hear about
speech recognition voodoo as true, check it out for yourself.

That applies to anything you hear from me, because I am just repeating
what I hear from the so-called experts.


I have found, much to my dismay, that those experts are sometimes dead
wrong.
Post by RW
Does anyone know who makes these OK headsets?
Not anymore. Nuance prefers to cover it up lately.


In the past, when ScanSoft owned Dragon, they left the makers name alone.

The "Andrea NC-91" anti noise headset was one such.

Another was the "Parrott VXI", which Dragon also used.



Generally I feel safe buying headsets and soundpods from eMicrophones,
because they cater to speech recognition users.


Other outfits cater to entertainers and hi-fi addicts, so I do not feel
as comfortable dealing with them.
Post by RW
And does anyone know of a desktop mic (they usually use
a 8" wand like stand) that has decent performance?
That means in a quiet environment.
Can't help you there. I found _one_ that works for me but the price
is astronomical. You might have to experiment a lot to find one you
like.
Post by RW
And does noise cancellation degrade audio quality in some way?
I expect it does. It works by canceling out background noise, which
sounds fine on the face of it.

But what happens when you are in a large closed room plagued by echoes?

Do not think it would work there.


True noise canceling is an exotic technology, if it is done right, the
benefits outweigh the drawbacks.

Look at how much money the "Bose" people are making with their
noise-canceling headsets for airline travelers.
Post by RW
Would it better to use an ordinary, quality mic
with a USB input in a very quiet environment?
I think it would, but where in the heck can you find a quiet environment?

You have to go with what is "practical". If an "approved" low cost
mic' does not degrade your accuracy enough to be noticeable, I see no
reason to buy a higher priced mic'.

I am not even too sure about the necessity to buy a $20 soundpod.

Perhaps it might pay off when dealing with a cramped MacBook where
everything is jammed together, but might not be needed with a wide open
desktop Mac where components are further apart.
Post by RW
Post by Mark Conrad
My particular USB soundpod is a Parrott VXI, bought from eMicrophones,
however I have used other brands of USB soundpods such as Andrea, and
they seem to work well also.
How would that compare to an iMic?
Do not know, someone would have to compare them.


I think you are in the same fix that I am, the need to actually try this
stuff to see if it works, which of course leads to wasting money at
times.

A good place to start is the lower priced stuff, just to get your feet
wet and see what the problems are.

That said, I still think the lowest priced Dragon "Standard" is a waste
of money, most people go up one notch and get the mid-range "Preferred"
version of Dragon, present version 9.00 of course.

Nuance wants $200, I have seen it as low as $118 at discount resellers.

I think Wes mentioned that he bought his for $160 at Amazon.

Assuming that, plus a $20 Parrott or Andrea USB soundpod from
eMicrophones, you would be out $180.


One important thing that has not been mentioned yet, setting up your
headset microphone from the damn Windows XP sound "Wizard".

If you screw up there, even a good microphone can be scuttled.

...if you are a guy who never had an encounter with Windows "Wizards",
lots of luck ;-)

Mark-
Mark Conrad
2007-05-02 20:58:07 UTC
Permalink
RW had questions about microphones, USB soundcard adapters, etc.





Corrections on Mark's previous post -

Whoops, misleading comment on my last post.

I blamed everything on the Windows XP audio setup Wizard.

Actually, audio settings on 3 things have to be set up:

1) Windows XP OS audio
2) Dragon audio
3) Mac OS X audio

For example, running OSX/Parallels/XP Pro/Dragon I was getting hardly
no audio out during playback. Had everything set correctly on XP Pro
and Dragon, but forgot to turn the volume up on OS X.


Another important thing, you showed interest in having a mic' work at a
few inches distant.

In that case, assuming you have a microphone that is capable of that,
you still have to jump through a few hoops.

The microphone setup Wizard in Dragon itself has some text that has to
be spoken in during initial audio mic' setup.

The trick is to speak that text in at too low a volume, which will cause
Dragon to reject your mic' because it has too low an output.

Repeat the audio volume test using slightly louder speaking, until you
get Dragon to just barely pass the audio volume test.

This will insure that Dragon cranks up its internal audio gain settings,
thereby letting you use your mic' at greater distances.



The Dragon mic' test following the audio volume test is the "quality"
test.

You will have to speak somewhat louder in order to get Dragon to pass
that particular test.

If you have a lousy mic', or background noise is too loud, or your sound
card is picking up too much electrical interference, Dragon will fail
that quality test and will not allow you to use Dragon at all.

...until you fix whatever problem that is causing the low quality audio.

(change mic' or change to different external soundcard or speak louder
or decrease the distance between the mic' and your mouth or perform test
in a quieter location)


On the positive side, every free headset that came with Dragon has
passed the quality test for me. Every USB external
adapter/sound-card/soundpod I own has also passed that quality test.

Never used the iMac adapter you refer to so do not know if it will pass,
but it most likely will.

Mark-

Mark Conrad
2007-05-01 05:46:00 UTC
Permalink
This is a tough call, "Benefits" of using Dragon on a Mac!

There are a heck of a lot more drawbacks, than there are benefits.

Let's get on with it though, so I can close this thread and resume doing
something more productive than this useless yack-yack.



It gets down to why people convert speech to text in the first place.

If we could remember everything we thought, heard, felt, saw, smelled,
there would be no need to convert anything to text.

(unless you are a civil servant, where you are required to do useless
things)

Text would be useless and obsolete, $MicroSlop$ would go broke, because
a lot of their income relies on gobs of text.



So okay, let's narrow it down a little, and just consider what we hear.

We could record dolphin jabber, and convert that to text. hmmm

We could consider street noises, elephant infrasound, insect ultrasound,
and try to convert _them_ to text. hmm, hmmm

Nothing there that a rational person would do. Perhaps a research
scientist would tackle some of the above projects, but then a lot of the
research scientists that I communicate with are not all that rational.

We _could_ convert our friend's and stranger's voices to text, but at
the present time this is extremely difficult, because speech software is
not that far advanced yet.

We could even train our speech software to recognize the synthetic
speech of our Macs, just for a useless fun project.




Being pragmatic, we are left with the unappetizing prospect of
converting all of our own speech to text, for whatever reasons.

Remember the main reason we convert our own speech to text, namely
because our memory is far from perfect.

Without referring to text to jog our memory, very nasty things could
happen to us.



So let's make an attempt to create a list of _practical_ reasons to
use speech app's.

Feel free to add to this list, if you can think of anything that can not
be solved by:
A) quickly scribbling a note
B) using a digital recorder
C) hunt and peck typing something


1) We just thought of a neat way to make a million dollars.
It is a long convoluted way. We _could_ record it on
a digital voice recorder, however we would really prefer
to have it as a text document, so we could process it
for obvious flaws that would otherwise be difficult to
spot from a voice recording.

2) We have a tiny digital voice recorder that is small enough
to fit in a man's shirt pocket. We see something important,
so we jabber into our recorder, speaking in all small yet very
important and critical details of what we are experiencing.
Dragon can convert every small detail of our observations
to text, so that NO important verbal details are "forgotten"
or worse yet, misinterpreted later.
I have actually done this with Dragon, the conversion of
speech to text is excellent, provided an "approved" type
of digital recorder is used.

3) We have hazy poorly formed thoughts about how to do a large
project. If those half-vast thoughts were in a text document,
we could easily spot ways of improving our ideas.

4) There are nasty "gotchas" involved in configuring software
app's, things that are not ITFM, things that have to be
learned AND recorded, else they will come back to bite you
if you ever need to reinstall the app' from scratch later.
I used to painfully write "gotcha" lists for each app',
now I can just flip on Dragon, easily record all the gotchas
with my voice.

5) We have manual skills that are largely automatic, leaving our
minds free to speak about what we are doing, therefore accurately
record what we are doing with our hands.
Seven years ago, my doctor was doing just that, using the
then current version of medical Dragon. I was conscious at the
time, so heard him describe the difficulty he was running into
during the operation, something about an artery on the back side
of my heart being too ulcerated to allow installing a stent.
That text record allows my present cardiac surgeon to more
accurately determine a course of action, when I pop in there
a few days from now.

6) We are too lazy to scribble or type, prefer to lay back in our
recliner chair listening to TV with our ear buds, dictating great
thoughts periodically into a nearby microphone, which is hooked up
to our Macs to convert our babbling to text.


All the above crap could have been dictated and sent to this NG while I
was in my easy chair watching the Pink Panther on TV.

All done from a Mac/Parallels/Window-XP/Dragon/dialup lash-up.

Mark-
Mark Conrad
2007-05-01 16:34:06 UTC
Permalink
Negatives -
*********
I want to get these out of the way first, so I can wrap this up on a
positive note.

I will not knock the EXTREMELY high price of Dragon too much, because it
takes a bunch of money to create and maintain such software.

That said, Nuance has really jacked up the price in recent years, with
very little to show in the way of improving the software itself.

Often their newest software can be bought from resellers at great
savings, so shop around before buying. I seldom buy directly from
Nuance.

Nuance is a big heartless international money-grubbing organization, who
is prone to greatly exaggerate the benefits of their software.

They have no compassion whatever.

They make a LOT of money because they cater to executives of large
companies who are often clueless about the real-world merits of speech
recognition software.

They fleece even more money from gullible professionals like doctors and
lawyers, who are led to believe that the software will save them a ton
of money.

Some of the software features are essentially useless and redundant,
such as the "roaming" feature that I have yet to find a practical use
for.

Like most Windows software, you have to give your first-born child to
Nuance in order to "activate" Dragon.

Nuance website really sucks, very hard to navigate. If a Mac user
stumbles in, he is steered towards an obsolete product named ViaVoice.



Previous owner of Dragon was an outfit named ScanSoft, much better as
far as us personal computer users were concerned.

Prior to ScanSoft buying it, the software was called Dragon Dictate,
created by a small startup company.

Speech recognition software is VERY DIFFICULT to configure and train for
maximum accuracy and speed, which has been found out the hard way by
some Mac users who tried it, and were disappointed at the results.

A lot of misinformation and voodoo is associated with the technology.

Lastly, it is a Major Undertaking to find "practical" uses, to justify
all the money and time involved in getting up to speed.



Positives -
*********
We live in the new Intel Mac era, where we can use a lot of $Microsoft$
software directly on our Macs if we are so inclined.

If we consider a Windows app' to be better than anything similar for the
Mac, we can use that PC software on our Macs.


Among other benefits, this tends to force Mac developers to field
software that is more competitive to the PC version of the same software.

Once the software and accessories are bought, configured, and trained,
we can chug along for years reaping the benefits of the software.

Real progress is slow when it comes to speech software, so there is
little worry about your software becoming obsolete overnight.

Speech-to-text has its place in this world, enabling us to do stuff we
could not otherwise do. It is just a tool, but sometimes that tool is
needed.

There is something magic about talking to your Mac, then seeing the
words appear almost flawlessly on the screen.

Even the few flaws can be fully corrected by speech alone.

The resulting large text document can be moved to the Mac side, by voice
of course.

On the Mac side, the text formatting can be handled by iListen, which
although not nearly as good as Dragon for dictation, is adequate for
such text formatting and layout chores.

An even easier way would be to do all text formatting and layout in
"Word" on the Windows side, then export the finished result to the Mac
side.

Dragon Pro can press keyboard keys and key combinations, all by voice
control.

Dragon Pro can do mouse clicks, double clicks, drags, cursor control,
edit images in photoshop, run databases, spreadsheets, send emails and
faxes etc., all by voice.

...often the finished Photoshop image, database, spreadsheet, can be
exported to the Mac side.


Due to the multitasking of the Mac, Dragon _should_ be able to start
Timbuktu, install TB2 on a distant Mac, control the distant Mac and the
entire distant house if the house has automation installed, such as its
sprinklers, security, fire suppression, appliances.

...all by voice, of course.

I have never actually tried this, so take this with a grain of salt.


Okay, let's get a little way out, using our imagination.

Combining voice control with artificial intelligence, we can yell at our
robots and order them to stack objects like balls, cones, odd shaped
boxes.

We can start and control heavy machinery by voice, keeping detailed text
records of all operations.

We can instruct an intern medical student how to do an operation,
keeping detailed text records as we progress.

Better yet, replace the intern with an android.

We can talk to those Japanese androids directly, to get them to
completely replace human workers on a construction site.
(automatically keeping detailed text records of our "orders")

Japanese are already having moderate success with these automated
construction workers.


We can start a very large company, name it "MacInSoft", populated by
androids that we yell at periodically.

Some of those androids would be software programmers, heavily laden with
built in artificial intelligence, to completely replace human
programmers.

Their software of course would be bug free, because their built in
computers are not capable of making stupid human mistakes.

End of post, end of thread, gotta take my med's because my brain is
over-heating.

Mark-
Davoud
2007-05-01 19:34:05 UTC
Permalink
Post by Mark Conrad
I will not knock the EXTREMELY high price of Dragon too much, because it
takes a bunch of money to create and maintain such software.
If one isn't an attorney or a physician DNS doesn't cost much. I paid
$160 at Amazon for"Dragon NaturallySpeaking 9 Preferred Speech
Recognition."
Post by Mark Conrad
That said, Nuance has really jacked up the price in recent years, with
very little to show in the way of improving the software itself.
Version 9 included *no* added features, only improvements.
Post by Mark Conrad
Often their newest software can be bought from resellers at great
savings, so shop around before buying. I seldom buy directly from
Nuance.
Isn't that nearly always the case? Doesn't one generally buy from the
least expensive reputable dealer, and isn't that dealer nearly always
someone other than the manufacturer?
Post by Mark Conrad
They fleece even more money from gullible professionals like doctors and
lawyers, who are led to believe that the software will save them a ton
of money.
Physicians love DNS. They say it saves them money because it saves them
time. There are people who are convinced that time is money.
Post by Mark Conrad
Like most Windows software, you have to give your first-born child to
Nuance in order to "activate" Dragon.
?? I activated Dragon in late March of this year, but I don't remember
anything about the process. If it had been difficult I would have
remembered. I'm racking my brain. As nearly as I can recall, activation
required a mouse click. I have a wide variety of Windows software; the
great majority of it did not require activation, while the software
that did require activation required a click of the mouse.
Post by Mark Conrad
Nuance website really sucks, very hard to navigate. If a Mac user
stumbles in, he is steered towards an obsolete product named ViaVoice.
Er, uh, Nuance doesn't make a product for the Mac. What do you expect
them to do, direct you to a make-believe page with DNS for Mac on it?
Post by Mark Conrad
Speech recognition software is VERY DIFFICULT to configure and train for
maximum accuracy and speed,
Not so. If one can read simple English sentences and manipulate a mouse
and keyboard one can train DNS.
Post by Mark Conrad
which has been found out the hard way by
some Mac users who tried it, and were disappointed at the results.
Ah, that was iListen. They're working hard to improve iListen, but it
isn't even close to DNS in accuracy at the moment.
Post by Mark Conrad
A lot of misinformation and voodoo is associated with the technology.
Your post doesn't help much in this regard!
Post by Mark Conrad
Lastly, it is a Major Undertaking to find "practical" uses, to justify
all the money and time involved in getting up to speed.
Whether speech-recognition software is worthwhile is for each
individual to decide for her or himself.

Davoud
--
usenet *at* davidillig dawt com
Mark Conrad
2007-05-01 22:27:54 UTC
Permalink
Post by Davoud
Post by Mark Conrad
Like most Windows software, you have to give your first-born child to
Nuance in order to "activate" Dragon.
?? I activated Dragon in late March of this year, but I don't remember
anything about the process. If it had been difficult I would have
remembered. I'm racking my brain. As nearly as I can recall, activation
required a mouse click.
See, it was so terrifying that your memory of it was expunged. ;-)


Good Grief, a real live Mac user of Dragon. Congratulations!

You probably have a better view of how useful Dragon can be to the
"ordinary Mac user". My biases towards/against Dragon have built up
over the years.

Original Dragon has changed owners more often than I change my shorts.

I forgot to mention that an outfit named "Lemout & Hauspie" owned Dragon
between "ScanSoft" and the present "Nuance".

These are BIG international speech outfits, L & H for example was based
at "Flanders Languages Valley" in Ieper, Belgium.
(something like our Silicon Valley, only for speech software)
Post by Davoud
Post by Mark Conrad
which has been found out the hard way by
some Mac users who tried it, and were disappointed at the results.
Ah, that was iListen. They're working hard to improve iListen, but it
isn't even close to DNS in accuracy at the moment.
Thanks, I hesitated to say that in so many words because people _know_
I am heavily biased in favor of Dragon.

It is the gosh awful truth, however.

iListen _could_ leapfrog Dragon, if they tied in Artificial
Intelligence techniques to the software.

The resulting app' would be huge, would need to run on only the very
fastest Macs, and would cost a lot more than Dragon does now.

No hobby market for such a product, only deep pocket businesses would be
able to afford it.


For here and now, the "Preferred" software version 9 you have is ideal
for Mac users; I spotted a price yesterday of $118 at Amazon, I think.

(Nuance wants $200)

Once Mac users get hooked by Dragon, some of them will get the expensive
Professional version.

I myself am poor as a church mouse, however I enjoy good hardware and
software, besides I think I will not live long enough to really suffer
from my forthcoming poverty.<g>


One question, how much does it bother you to have to leave Dragon
Preferred fired up until you finish the correction operation?

Big Deal or minor annoyance?

Mark-
Davoud
2007-05-02 13:22:16 UTC
Permalink
Post by Mark Conrad
One question, how much does it bother you to have to leave Dragon
Preferred fired up until you finish the correction operation?
Big Deal or minor annoyance?
II'm not sure what you mean. No big deal, I guess, because it's a step
forward that improves performance.

Davoud
--
usenet *at* davidillig dawt com
Mark Conrad
2007-05-02 17:54:24 UTC
Permalink
Post by Davoud
Post by Mark Conrad
One question, how much does it bother you to have to leave Dragon
Preferred fired up until you finish the correction operation?
Big Deal or minor annoyance?
I'm not sure what you mean.
With the "Preferred" version of Dragon that you have, if you exit
Dragon, the original audio is flushed out, no longer available.

That means if you postpone correcting your text by shutting down Dragon
_before_ correcting your text, you have no way of knowing whether or
not you mispronounced a word.

If you accidentally mispronounced:

"Sue sells she shells by the sea shore"

...and try to correct "she" to be "sea", it tends to screw up the Dragon
vocabulary such that every time you later say "she", Dragon messes up
and grinds out "sea" instead.


With the original audio intact, you never correct a mispronounced word.


(or at least you _should_ never correct a mispronounced word)
Post by Davoud
Post by Mark Conrad
Big Deal or minor annoyance?
...<clip>... No big deal, I guess, because it's a step
forward that improves performance.
Great, that saves you $700 by not having to get to Pro version, which
saves the original audio between Dragon shutdowns. :)

Mark-
Wes Groleau
2007-05-02 00:59:51 UTC
Permalink
Post by Mark Conrad
Dragon Pro can do mouse clicks, double clicks, drags, cursor control,
...., run databases, spreadsheets, send emails and
faxes etc., all by voice.
Which is why I and a friend set it up for another friend
with limited motor control due to Parkinson's.
Post by Mark Conrad
Some of those androids would be software programmers, heavily laden with
built in artificial intelligence, to completely replace human
programmers.
Their software of course would be bug free, because their built in
computers are not capable of making stupid human mistakes.
Four thousand processors plus one terabyte of memory
equals half a mouse-brain with one-tenth the speed.

http://news.bbc.co.uk/2/hi/technology/6600965.stm

I don't think we'll be seeing your androids very soon.
--
Wes Groleau
"What you see and hear depends a good deal on where you are standing;
it also depends on what kind of person you are."
-- C.S.Lewis
Wes Groleau
2007-05-02 00:50:38 UTC
Permalink
Post by Mark Conrad
Seven years ago, my doctor was doing just that, using the
then current version of medical Dragon. I was conscious at the
time, so heard him describe the difficulty he was running into
during the operation, something about an artery on the back side
of my heart being too ulcerated to allow installing a stent.
Even with Dragon, this demands a human to listen carefully to the
recording and compare it to the text. Serious liability is risked
by skimping on this step. A relatively harmless example: My wife's
medical records say she had a lymph node removed. It was actually
a "lipoma" (a tumor-like glob of fat just under the skin).
Post by Mark Conrad
All the above crap could have been dictated and sent to this NG while I
was in my easy chair watching the Pink Panther on TV.
And we would wonder why so many

"Clouseau!"

were sprinkled through your post.
--
Wes Groleau

http://Ideas.Lang-Learn.us/
For lovers of language and learning
Continue reading on narkive:
Loading...