2002 Conference Proceedings

Go to previous article 
Go to next article 
Return to 2002 Table of Contents


VOICE RECOGNITION TRAINING: 60 MINUTES TO SUCCESS

Doug Bowes
SET-BC
Special Education Technology-British Columbia
2260 Victor Street
Victoria BC
V8R 4C5
(250) 595 7511
dbowes@setbc.org

People buy voice recognition software for a variety of reasons. Some simply want the convenience of using their voice to type. Others, who can't use their hands or have a literacy-based disability, have a deeper need. They need voice recognition to be able to put their thoughts down on paper. Many are likely to fail because of a poor start.

This session will focus on getting your client up and running with a good rate of dictation and high accuracy. Featured will be Dragon Naturally Speaking and Kurzweil Voice. The methods and adaptations will apply to other programs as well (Via Voice Dragon Dictate).

EQUIPMENT

Before you start training, you need to know that you have the right computer to run your software. A poor software/hardware match can stop you in your tracks. Preparation will include hardware compatibility checks, environment and sound level modifications, and voice recognition software settings. Even with a high end Pentium III, plenty of RAM and a 16 bit sound card you may not have a good enough match. Alternatives to the on board sound card will be presented as well as after market microphones.

ENVIROMENT

Trainers need to be aware that environments change both the location and by circumstances. Changes to ambient noise levels and their affect on speech patterns have to be taken into consideration. Recalibration and shorter dictating periods are often in order.

Noise canceling microphones are only part of the equation. You must consider how changing sound environments affect the user and accommodate accordingly.

TRAINING

The most critical part of the process is training. It should be quick and give the user the ability to write independently by the end of the first session. The best way to achieve this is to model a speech pattern that is measured, deliberate and where each word has a distinct beginning and ending. Start by reading a training passage (using a measured deliberate rate where each word has a distinct beginning and ending).

When the passage is complete, demonstrate an effective dictation method, correction, editing formatting and navigation. Keep it basic. Too much information and your client may be overwhelmed. Turn the microphone over to your client and create a profile for him. Use the same training passage as you used. Make sure that that you moderate the rate and diction as the reading progresses. Encourage the user to speak slowly and clearly.

Use the read back feature to illustrate why the computer has misinterpreted speech

Remember the following

1. Each word should have a distinct beginning and ending
2. Guide the vocabulary and dictation
3. Don't let mistakes compound and keep control of the dictating environment.
4. Test misrecognitions to learn their cause
5. Follow up within two days

CONSIDERATIONS

There are training adaptations required for users with low literacy levels and reading difficulties. The level of functional literacy will determine which product is used, how much support is required and whether voice recognition is used as a writing tool or a literacy tool. Your expectations should be reasonable as voice recognition is not a panacea.

FOLLOW UP

The first follow up session should be with in two days. Be prepared to note and correct poor dictation and correction practices. Make sure that poor practices are corrected and that the client knows how to increase speed and accuracy.

SUMMARY

A modified training approach along with optimal equipment configuration can have a user up and dictating within 60 minutes.


Go to previous article 
Go to next article 
Return to 2002 Table of Contents 
Return to Table of Proceedings


Reprinted with author(s) permission. Author(s) retain copyright.