Voice Recognition Works. Really!

I have been experimenting with voice recognition rate recently, with pretty nice results.  Over the years, I've tried IBM ViaVoice, Dragon 6, and Dragon NaturallySpeaking 9.  My first experience with voice recognition, I had to speak like a robot and the computer translated my sayings into a very odd but beautiful poetry.  Although it was very satisfying on an emotional level, it wasn't very useful for me.

With Dragon NaturallySpeaking 9, I have been pleasantly surprised.  Although voice recognition has for years promised a 99% accuracy rate, I am starting to see fewer and fewer errors.  With my current setup, I have had maybe one error every other sentence.  This is not what I would consider to be a 99% accuracy rate, but it is acceptable in some senses for producing a rough draft.

The thought has been lingering in the back of my head, however, that maybe what I really needed was a new microphone.  So it was with some sense of glee when I spotted a new headset on the desk at work.  My colleagues have purchased a Microsoft LiveChat 3000 headset to do video casting, which they graciously let me tear into with my greedy little fingers and test first with the voice recognition.

From the moment that I heard the playback on my audio test, I was addicted.  The sound was crystal sharp, clear, precise,... dare I say "beautiful?" Okay, maybe not beautiful, but the noise reduction made a huge difference, and I found that I can speak at a normal speed and have every word appear on screen as it should be.  Whereas I had an error every other sentence before, now I had maybe one error in the paragraph.

So what did I do?  I couldn't take this headset home with me.  My colleagues would have been very upset.  But I can no longer use that substandard, el cheapo headset.  I was hooked.  I wanted perfection.

I got online and did a little research.  It's hard to find many useful sites on voice recognition.  I think most people don't take it seriously.  A recent Slashdot posting poked fun at voice recognition, thanks to a horrendous demonstration given by Microsoft.  I found a few sites, which I will reference later, that reviewed microphones and software.  But there aren't a lot of community sites out there that I can find.

According to the sites I did find, microphones by Sennheiser, Plantronics, and Andrea are all recommended.  Even with these recommendations, they tend to be one or two years out of date and the microphones are off the market.  Or, in the back of your mind, you're wondering if new technology has pushed the price down of a $200 microphone to $50.

I found nothing about voice recognition and the Microsoft headset. With the sites recommending microphones that cost over $200, it was very hard for me to limit myself to the Microsoft headset, which cost only $28 at Amazon.  However, I didn't want to spend more than $50.

I checked the time, and found that I had only 20 minutes until Home Depot closed.  Although I'm regularly not an impulse shopper, I really was hooked on getting this tonight.  According to their website, Home Depot has the Microsoft headset so I decided to drive there to pick it up.  Incidentally, it is very frustrating when a company advertises something online that it does not have at the store. I just want to say that, because once I arrived I was very frustrated that Home Depot in fact did not have the object of my desire.

Unwilling to go home empty-handed back to my cheapo headset, I decided to take a risk and buy the Plantronics 350 instead.  The Plantronics had good reviews and seemed like a reasonable bet.  It was also the same price.  If it didn't work out, I thought, it would be a nice gift for somebody at Christmas.

And what a fantastic decision that was to buy this device. Although I can't listen to myself like I could with the Microsoft headphones, the quality of the microphone is perfect.  When I say "Perfect," I mean perfect.  I think with a little training I should have that near 99% accuracy rate. I would be happy with a 90% accuracy rate.

By way of example, I will not change anything in the next two paragraphs.  This is unedited, exactly like I spoke it.  Which is very challenging for me to, because for the moment I write more clearly than I speak.  As you can see, in the previous sentence, there is one mistake but that is very easy for me to catch.  I also do have trouble organizing my thoughts clearly what I'm speaking.

A baseball game just finished. The Mariners completed another come-from-behind victory against the Baltimore Orioles.  Losing four to one in the seventh inning, the Mariners managed to score 3 runs to tie the game.  No, I was wrong they scored four runs to take the lead.  The Baltimore Orioles have some pretty bad relief pitching.

So in those two paragraphs, just a few minor mistakes.  Considering how easy they will be to catch on editing, I don't consider this to be a problem at all.  The important thing is, will I understand my writing when I am editing it?  Because if it is close to the original, even if there are a few mistakes, I can clean them out pretty quickly.

Again my hope is that I will be able to build my speaking speed to match that of my writing speed.  I know for most people it is the opposite that is true, but I think much more clearly what I'm writing and I'm talking.  It is, however, so nice to spare my wrists that I must master this.  Voice recognition is not the joke that it used to be.

For more information:

• Nuance (www.nuance.com) - Dragon Naturally Speaking 9
• Voice Recognition (www.voicerecognition.com/) - Check out the forum for a lot of useful information
• Plantronics Audio .350 - Bought it for $34.99 at Office Depot; as of today, it's $21.99 at Amazon.  Sigh.

Other facts:

Hardware: I'm using a Dell Inspiron 6000 with 1 GB of RAM and a 1.6 GHz processor. This computer is about 2 years old. It is also well known for having mic issues, if you use the mic input. I am using an iMic to convert the input to USB. It works very well. 

Software: Dragon Naturally Speaking 9 Preferred Edition