We haven't reviewed voice dictation software for several years. The last version of Dragon Naturally Speaking that we looked at was probably somewhere around version 6. Nuance, the company that acquired the Dragon software system, has recently released version 10 of the Dragon Naturally Speaking Professional series. It was received with great anticipation. We eagerly installed it on a test system, after which I sat there quietly thinking to myself, "what am I going to say about this product?" And that's when it hit me. 'Saying' is what it's all about with speech recognition software.
This minor epiphany led me to decide that I would give it the same test an average user would give it. I would ignore most of the instructions while maintaining high expectations for success. I installed it and immediately started dictating. To say the least, I was astounded. Accuracy was over 95% out-of-the-box. And to be completely honest, I did a few things that aren't necessarily recommended out-of-the-box.
I did follow all the basic quick start directions, which I would hope
every new user would do. This meant doing the basic recognition training. I
like to use an excerpt supplied by author Dave Barry (always humerous),
and then jump merrily into dictating a document. The level of accuracy
achieved in out-of-the-box voice recognition has improved dramatically
since last we looked at this market space.
The article you are reading
is being dictated in Dragon Naturally Speaking 10 Professional. With
the exception of formatting for the website, you are seeing exactly
what was dictated*.
*including corrections which were also done through the program
A Word About Hardware
You may have noticed that the heading above is in bold characters. This
was also, accomplished using simple commands embedded in the software.
But I digress.
Voice recognition software is notorious for requiring a lot of
processing power. The test machine being used incorporates a quad core
AMD processor and 2 GB of RAM on a Windows XP platform. This is not
because the software requires that much power. According to
specifications it only requires a 1 GHz processor and 512 MB of RAM in
a Windows XP environment. It's just that I'm impatient for the results
hence, the bigger processor. I will tell you that the text is showing
up virtually in real time.
Nuance package Dragon Naturally Speaking 10 with a nice headset/microphone combination from Andrea Labs. Andrea Labs has been
in the sound processing business for many years, and we have enjoyed
seeing their developments in sound envelope and sound beam technology.
To that end, I decided to go one better, and use their sound Max
Superbeam array microphone. I chose to do this not only because of the
phenomenal performance of this desktop field array microphone, but also
because I just don't like something stuck on my ear for any length of
time. With the field array mic, I can sit back and dictate with no
encumbrances. It also helps prevent me from looking like a complete
idiot when the phone rings and I try to answer it on the covered ear.
I ran the audio tests as part of the setup and the accuracy and sound
reproduction were excellent. I must stress that this is a critical
element of the speech recognition process. It is probably the best real
world example of GIGO (garbage in, garbage out). Nuance made an
excellent choice in choosing Andrea Labs to supply the headset/input
device.
It's Not What You Say
Dragon Naturally Speaking does a good job of incorporating the vast majority of
words and commands that are commonly used. The included vocabulary is
quite extensive and in version 10 they have greatly increased the
command and control capabilities. You can now easily initiate complex
commands to control programs in Microsoft Office products as well as
your browser, including the popular Mozilla Firefox. If you use Google
desktop, you can speak commands to search the files on your computer or
find something in your e-mails. If you're the adventurous type, you can
also script your own commands to alleviate the tedium of repetitive
tasks.
Certainly consistency is important with any speech recognition
software. Dragon Naturally Speaking gives you all the tools to train
and improve the speech recognition abilities of the software. As I
stated earlier, out-of-the-box accuracy is surprising, and with minimal
initial training, taking the time to make corrections in the document
as you develop it, you can improve accuracy to virtually 100%.

Like anything else in life, practice makes perfect. In the early days
of speech recognition, training was a true tedium. Dragon Naturally
Speaking 10 has eliminated that hurdle. With their improved speech
algorithms my fear of having to spend hours talking into a microphone
to get reasonable results has been eliminated.
A word of caution. No matter how focused you may be dictating a
letter, white paper, or simple e-mail.... life happens. Phones will
ring, people will walk in, talking, and extraneous noises will
interfere with the process. There's no way to avoid this, however, with
Dragon Naturally Speaking's error correction tools, cleaning up those
little annoyances takes but a moment.
Conclusions
Working with Dragon Naturally Speaking 10 is not only simple, it's addictive.
Being able to spew out words at a rate far superior to my typing
abilities makes me constantly looking for ways to apply the voice
recognition system to other applications I use daily. Fortunately for
me, Dragon Naturally Speaking has taken the time to set up the majority of commands I want
to use for my Office and browser applications. Nuance has proven their
mastery of speech recognition on the personal computer by getting it to
respond to my verbal requests and commands. Now if I could just get my
staff to listen the same way.
Dragon Naturally Speaking 10.1 Preferred-Boxed Shipment - $199.99
Dragon Naturally Speaking is also available for Medical and Legal professionals .
|
|
|