1999 Conference Proceedings

Go to previous article 
Go to next article 
Return to 1999 Conference Table of Contents


Getting it Right and Making it Work ! Selecting the Right Speech Input Writing Software for Users with Special Needs

Douglas R. Bowes
Special Education Technology-British Columbia (SET-BC)
2260 Victor Street
Victoria, B.C. V8R 4C5
voice/message: (250) 595-7511
fax (250) 595-7224
Internet: dbowes@setbc.org

Introduction

Presently there are more than twenty-five speech input programs for people who want to write with a computer using their voices. The purpose of this workshop is to help you select the best software/hardware combination to get the right software and make it work! It deals with selecting speech input writing software and computers for special needs and tips, tricks and training users to use speech input computing called Making it Work!

From a decade ago until the fall of 1998, speech input software and the associated computer to run it were a very expensive investment for schools or families choosing this approach to writing . Users found the interface was awkward and involved a great deal of training in order to master technology for speech input computing. In the beginning the software and computer could cost upward of 20 thousand dollars and take dozens of hours training before the user was even able to begin dictating text. Command and control features (those computer commands most often done with a mouse today) were limited or nonexistent. In short you needed a motivated user, a well trained support team and an abundance of time and money.

Today’s systems offer a short learning curve, a small price tag, and will operate on typical entry level computers. Speech input software is now mainstream software that will have a wider basis of support and be more easily understood by the general public than it’s predecessors. It is possible to purchase a complete speech recognition system (computer and software) for under a thousand dollars.

Getting the right computer

Speech input software needs a particular combination and configuration of computer hardware to be able to run.

This computer must have the proper sound card, a processor (CPU) that is fast enough and sufficient memory (RAM).

To run Kurzweil’s Voice Pad, Voice Pro, Voice Plus, Dragon Dictate, IBM’s VoiceType or SimplySpeaking you should use a Pentium 100Mhz with 16 MB RAM. Better performance will occur with more memory and faster processors.

Dragon’s NaturallySpeaking requires a minimum Pentium 133 with 32MB of RAM (up to 64 MB may be required to access all of it’s features and dictate directly into Microsoft Word) . If money allows a Pentium II is recommended as the processor.

Getting the right sound card

All current speech input software requires a 16 bit Sound Blaster (or compatible). Not all 16 bit sound cards are the same. Products for continuous speech such as ViaVoice and NaturallySpeaking are fussy about sound card compatibility. It would be best to test your software on any computer you intend to use for speech input computing (especially laptops). Sound card and software incompatibility is the most common reason for poor speech recognition in voice input computing.

Getting the right microphone

Most speech input software comes bundled with a noise canceling headset microphone. The software manufacturers ensure that they supply a microphone with good acoustical properties suited to their product, however many of these microphones suffer from design flaws that make them fragile or ill fitting. This is particularly important for young users If the headband keeps slipping or breaking the computer will have a hard time getting used to your speech. If you can’t find a headband that is suitable to your needs try the Telex Nomad ear mounted headset. This unit clips behind and over the ear and provides an inexpensive and comfortable alternative to many bundled microphones.

Getting the right software

Here are some thumbnail software recommendations for some types of special needs. Keep in mind that with enough time effort and support almost all software can be made to work for any user. The following are my recommendations to match needs to software with the least amount of effort. Your goal should be to have your user using the voice input software independently with in 2 - 3 weeks.

Young Children: Continuous input software ( such as ViaVoice & NaturallySpeaking) is not recommended for use by children (those with non-adult speech patterns).When speech input is required Kurzweil’s Voice line of programs best meet the needs for users under 10 years of age. Teachers and parents should examine the reasons for introducing speech input for users under 10. The most compelling reasons are when the quantity and quality of written work fall significantly below that of the user’s peers. Use of a giggle switch is highly recommended (see giggle switch below).

High written output (more than 200 words per day): For those adult users with the need for large quantities of written output, IBM’s ViaVoice or Dragons NaturallySpeaking would be the programs of choice. IBM’s SimplySpeaking is the best choice for children with high written output requirements and cannot use ViaVoice or NaturallySpeaking because they do not posses an adult speech model. Various versions of ViaVoice and NaturallySpeaking offer different voice directed command and control features for those who cannot use a mouse.

Speech differences, speech disorders, start with VoicePad (if the user can’t train VoicePad they are unlikely to succeed with other products). If command and control are issues Dragon Dictate is the best product. The more severe the speech difference the user has the less likely the user will succeed with products such as ViaVoice and NaturallySpeaking. N.B. A speech input program can become a powerful adjunct to a speech therapy program.

Voice disorders: People with voice disorders should consult with a speech language pathologist before commencing speech input computing. Those prone to vocal abuse or vocal nodules should consider continuous speech software such as NaturallySpeaking or ViaVoice.

Learning disabilities

None of the software packages presented are well suited for independent use by users with learning disabilities. The speech feedback features in SimplySpeaking, ViaVoice, and NaturallySpeaking (Deluxe and Preferred) is a useful feature for those users who would benefit from this type review . A major disadvantage in using continuous speech programs is the (1-3 seconds) time delay from the instant the user speaks the words and their appearance on the screen can be confusing and counter productive to the user with learning disabilities. The best results may be obtained by using a word by word approach using a discrete speech program such as Dragon Dictate or Kurzweil Voice.

I have had a great deal of success working with VoicePad (and other Kurzweil discrete speech products) and a modified problem solving approach to training and support. The fall of 1998 promises a new speech input product specially designed to support the writing and reading process for users with learning disabilities.

Making IT Work

If the user is a literate adult male with a broadcast standard dialect i.e. Peter Jennings need read no further. Others with speech , pitch, strong dialect differences or breathing and stamina problems will need to make modifications to equipment and training . Some of the strategies you may try include...

These and other training strategies will be offered to workshop participants.


Go to previous article 
Go to next article 
Return to 1999 Conference Table of Contents 
Return to Table of Proceedings


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