I've tried using Speech Recognition soft- and hard-ware more than a few
times over the last 35 years.
Hi Name a rooni,
Don't know about the last 35 years but I can honestly say it has only become
a serious and viable proposition in the last four. The reason I know this?
Because in 2004 I had a near fatal car accident, broke my neck, and became
paralysed thus ending up a quadriplegic.
I eventually ended up in the rehabilitation section in the spinal wing of
the hospital and spent five months in there. There, they have a whole room
filled with computers dedicated for patients being rehabilitated. For
quadriplegics as me, there were only a couple of choices for obvious
reasons.
Dragon at that time was very frustrating and did not work all that well.
For a quadriplegic, this was very frustrating. The training took forever
and even spending days on it, the dictation that came on-screen was full of
mistakes and looked more like Chinese rather than English. For me, it was
very soul-destroying and sent me even more into depression.
The only other alternative which name escapes me was one where you put a
tiny, circled piece of material that you put on the end of your nose or
forehead that was supposed to stay there and stick being the special
material it was. When the program was all plugged in, there was a tiny laser
beam between the ends of your nose or forehead depending on where the circle
was stuck that projected down to the keyboard. The idea was you just
pointed this laser beam towards the letters on the keyboard by moving your
head appropriately and the letters would come up on the screen.
At first, it seemed a more viable proposition than Dragon because as I told
you, Dragon at that time was next to useless. The trouble was after a
while, moving your head around to point to all the different letters on the
keyboard became very tedious and my neck started to ache after a while. I
also found that after you used the program a few times, the cursor became
stuck to the screen and you were forever nodding your head like a rabid
nodding dog in the back of a car to trying to move it thus making your neck
hurt even more!
I eventually (like most disabled people then) gave up on both applications
and slithered further down into depression.
After I came out of hospital, it wasn't till 2007 that I applied for a grant
so that I could buy a new laptop computer because this was the only
practical way of having a computer next to me really. I had also heard
about another speech application that was built in to Vista, (being the
operating system at the time) called WSR short for 'Windows speech
recognition'. I did not hold up much hope after my experiences in hospital
but it was more intrigue than anything and what did I have to lose anyway.
Bearing in mind this was only about three years after experiencing the
failure of speech recognition in 2004 with Dragon. Well I can tell you, I
was so surprised with this speech app because to some extent, it worked
quite well and at least you could string a sentence together unlike the days
in 2004. Bearing in mind, I had just started using it and I was not using a
decent microphone (only the integral one in the laptop itself), I began to
use it more and suddenly became better at it that then gave me the bug. I
found myself writing e-mails and other stuff reasonably easy and then
decided to campaign to other charities to get the necessary tools to make it
even better. I eventually got myself a better computer and that choice of
decent microphones. (A decent microphone is imperative if you want to
pursue speech recognition seriously).
Anyway, fast-forward two years with plenty of practice and training (and
time) I find it second nature as you can see. I have no choice because but
to use it, as I haven't the luxury of simply reverting to the keyboard if I
fancy a change or a mistake comes up. So it's a way of life of me now.
After becoming proficient with WSR, (this is what you should start with if
you want to learn and this is free now anyway in any Windows operating
system) then I campaigned for the funds to buy Dragon (because it is rather
expensive) because I heard it had vastly improved since my days in hospital.
I eventually got my funding and the speech expert who guided me through my
first couple of days of training, insisted that I have nothing less than
Dragon Professional because with this version you can make your own macros
(commands that tell speech recognition what to do on your computer). To buy
this version though will set you back around £ 500.00 ($1200-1300) which is
not exactly cheap. That is why I suggest if you are serious and have the
time to dabble in this software, to start with WSR first because it is free
anyway.
Since I have had Dragon (since November last year) I have found it is a
great asset for someone as me and the dictation side is more accurate than
WSR at the moment but then it's been around a great deal longer that WSR.
I use both, because I find WSR quicker on the command side (for getting
around the computer, changing programs, surfing the web etc.) And Dragon
for dictating, especially long documents.
I'd already been a proficient touch-typist for some time already, typing
groups of letters and numbers (not recognizable words) at up to 80wpm (no
prizes for guessing >where, but if I told you I would have to kill
something).
I bet ;-).
The average typist can type 40 words per minute but once you've mastered the
art of speech recognition, you can easily treble that to about 120 words per
minute without any mistakes provided you speak clearly and concisely and
enunciate every word properly. This is what WSR and Dragon recommend.
I have got both speech apps up to 150 words per minute before but that is
speaking very fast adding the risk of bringing up mistakes so there is no
need to speak like a that.
Anybody who tells you they can speak faster than this and expect to bring up
zero mistakes using speech recognition is lying. Try it yourself without a
speech app by reading from a newspaper or book and time yourself for
one-minute then count how many words you actually spoke. Bearing in mind
that they all would have to be spoken clearly and concisely to have a chance
to come out on screen with zero mistakes if you were using a speech app.
Look at this link.
http://www.dragonvoicerecognition.com/dragon-professional.cfm
Look under the heading: "Features for Dragon NaturallySpeaking professional
solutions."
.... And then just underneath the subheading: "Up to 99% Accurate and Three
Times Faster than Typing.
It tells you how many words per minute Dragon reckons most people can speak
in reality.
Still, I could get a document typed from copy or my own thoughts in a
fraction of the time; even if I slowed down to ensure that it was 100%
perfect and formatted correctly. The danger was thinking too fast for my
fingers. I had to consciously take time to form the sequence of words from
my thoughts. Rarely happened though because what I usually typed came from
outside my head, and it was a case of type it or lose it...
This is also the case with speech recognition in some respects. Copying
from a script such as a magazine or book for example is a good test for
dictation. When you have to think about what you are going to say before
putting it on screen is a different kettle of fish and will inevitably slow
you down a bit.
Leap forward to the 80's; I first tried SR software. The time it took me to
educate the software to recognize my voice was inordinately long. Still
typing, I gave up after less than a month; could still type a document in a
fraction of the time (computer keyboards are amazing compared with
teletype) even if I slowed down.
I quite believe you because speech recognition in 2004 was atrocious and not
worth the effort so I cannot imagine what it must have been like in the 80s.
No wonder you gave up.
The only problem was thinking too fast for my fingers; I had to consciously
form the sequence of words from my thoughts.
It's the same with speech recognition; you have to know exactly what you're
going to say before dictating it on screen otherwise you will find yourself
putting down stuff you do not want. It pays to pause a little before
deciding what your next set of words should be.
1995: tried it again. No signs of remarkable improvement. The training
period was still too long to be worthwhile.
Yep! Wasn't worth a jot until at least 2006.
2003: tried it again. Ditto the above.
Agreed.
Do I have the time to train a piece of software? Nope. If I was doing
that, I wouldn't be communicating, or partying.
I would have thought the same before I had my accident.
I had my own business and was a busy bee with no time to learn something I
had the need for. Using the keyboard was enough for me anyway. Did not need
any speech recognition then. Now I have all the time in the world so I had
spent it wisely and productively by learning how to use speech recognition
properly. Besides, I have no other option if I wish to put my thoughts to
the screen or write a letter or e-mail etc.
The years you are mentioning also took time with training and speech
recognition didn't work that well and wasn't worth it anyway as I explained
above until at least 2006.
Would I use one if that training was made much shorter? Yup. Can't afford
an amanuensis.
Today, I think you would be mildly surprised. I believe they have finally
got their act together. Training is relatively short now from the bad old
days and after that, you can virtually start dictating straight away.
Of course, it takes time to learn the commands and punctuation and if you
haven't got that time, you may as well forget it. But Rome was not built in
a day and you do not have to learn all of this in one hit. As I've said
before, I had no choice so I had to learn but I did have time, which is a
great help if you want to master speech recognition. With it comes great
rewards and you will eventually find you can dictate including all
punctuation with relatively low mistakes at three times the speed of an
average typist.
That is provided you have the proper equipment e.g. a decent
microphone/computer etc. That is essential for you to succeed in this
endeavour ;-).
andy t
