Voice Recognition With An Aussie Accent

With the release of Windows Vista, I thought that I would try out speech recognition again to see how far it had come.

As an Australian, speech recognition for me is a hit and miss affair. I have enough trouble getting ordinary Americans to understand how I pronounce words more complex than G’Day, Crikey! And That’s not a knife!  A friend of mine is sure that I manage to get at least 15 vowels out when I pronounce “Coca Cola”.

Australians have long since learned that the easiest way to get speech recognition to work is to try to imitate the American accent. After being reliably informed that my attempt at a Texas Twang sounded like someone was torturing a cat, I’ve switched to an imitation that sounds like Forrest Gump crossed with Paul Hogan.

So, in general, the whole speech recognition thing hasn’t been a resounding success.

Vista includes speech recognition out of the box. It was around in Microsoft Office and Tablet PC edition, but Vista marks the first time that Microsoft has been confident enough to include it as an operating system feature. Except that for all the publicity about Vista, the first I knew that speech recognition was available was when I noticed its icon sitting in the Vista control panel.

The algorithm appears to be adaptive. If you correct Vista enough times, it works out your pronunciation. You say: “Correct wrong-word” and a correction dialog appears. Hopefully the word you want is present. If it isn’t, you try and say it again. If it still doesn’t appear on the list you tell the computer: Spell it and from there you can spell each letter of the word to get it right.

It is about here that I really hit a problem. To the American ear, the Australian pronunciation of the letters A, E or I sound pretty much the same. They do to Windows Vista as well. There isn’t an option to manually enter text via keyboard or Tablet pen so I got a little stuck. I had to try every pronunciation of these letters in the book before I was able to get Vista to figure out what the heck I was on about.

Now that my Tablet has been trained, things are a little better. I’m not sure that it is because Vista understands what I’m saying, or that I’ve changed my pronunciation after shouting I, I, I, not A, I, No I said I, not E, I, I, I, Damn, Delete Damn, I for about half an hour at the screen.

Don’t get me started on letter H (Ache, Aike, Ike, Aych …)