Wise Owl Opportunities

Imagine a computer with no keyboard. Imagine no typing. Thanks to new versions of speech recognition software, this concept is much closer to reality then we may think.

Microsoft's new Office XP offers voice recognition that lets you talk your way through your activities. Other products such as IBM's Via Voice & Dragon Naturally Speaking have been on the market for a while. These products incorporate Speech Recognition by using spoken words to dictate text and commands to navigate menus and make changes.

Too good to be true? Well, currently the manufacturers lead you to believe that it takes just five minutes to get started. In reality it takes more than just a little training to get the hang of voice recognition - figure on a couple weeks.

However, I was up and running in a couple of hours in Word utilizing the voice recognition as a compliment to the keyboard and mouse. Now, remember I have a strong computer background, speak very clearly, and have worked with voice recognition programs in the past. The truth is Voice Recognition does not work immediately out of the box. You'll need to train it. Then you'll need to train some more.

In essence you need to allow the program to become familiar with your voice. You'll spend time dictating stories so the program can get acquainted with your intonation and pronunciation. If you're new to dictation, there's a big learning curve associated with this way of working. How well you annunciate and how well you follow the programs guidelines will determine just how much training is required.

So how does it work? When you talk into the microphone the computer will hear your speech as a continuous stream of sounds. To match these sounds to words, applications rely on two sources of data: acoustic and language.

Acoustic data matches your sound patterns that make up different words and chooses the words that most closely match what it heard. Since no two people sound exactly alike, the programs do a much better job when it knows something about your pronunciation.

Language data relies on fact. Consider two phrases: "pizza delivery boy" and "Pete's a delivery boy." When spoken, they sound almost exactly alike. The phrase "pizza delivery boy" is more common than "Pete's a delivery boy," so the program would favor this phrase over the other.

Here are some tips when your computer is in training:
  1. Position your microphone correctly - experiment with moving the microphone for best clarity.
  2. Speak properly and pronounce each word clearly. Remember, your computer is not a person; it won't       help to shout, or slow your speech to a crawl.
  3. Speak in phrases - Don't say "Dear Lauri" one word at a time, the computer might type "Deer" or       "Gear". The whole phrase ("Dear Lauri") allows the program to use context to determine the most       likely "Dear."
  4. Work in a Quiet Environment - noises are distractions for the computer.

So, if typing slows you down and are anxious to join in on the computer fun … start talking!