Sit down at a computer, put on a lightweight headset, talk into the microphone and lo, your words, spelled correctly, come up on the screen. That, in essence, is what voice recognition software is all about. Surely, the special needs teacher will say, this is the way forward for pupils with dyslexia or other physical or cognitive problems with the written word.

At Bilston FE College, in the West Midlands, Robin and Deana Korwin, a husband and wife team, run a heavily oversubscribed Saturday morning family dyslexia support group. Deana is the college’s family learning manager who started using the software with adult dyslexics as soon as she found out about it. “We began to see wonderful benefits. Their confidence grew, and one lady who could never put a sentence together now does wonderful poems,” she says.

About 10 families attend the Saturday morning sessions - usually a child with his or her mother. There is a “whole-class” session where general topics are covered , but the families come and go from this group for half-hour sessons next door on three computers, where they are individually supervised by Robin Korwin or a colleague.

I watched several children using Dragon Dictate, which accepts words dictated separately, with tiny gaps between. Running with it was Keystone, a program that reads aloud what the child has just dictated. The child, therefore, says the word, sees it, and hears it read back.

The potential is already obvious. The children see the eventual benefits, and are willing to work hard and long at the difficult early stages, when the system has to get to know the user’s voice. It never mis-spells a word, or produces incorrect grammar, but it may well put up on the screen something very different from what you said.

Twelve year old Matthew found it quite funny when he dictated the word “frown” only to see “Ferrari” appear. The correction process has to be properly done, using on-screen menus and voice prompts so that the software will never make the same error again - it will know that when Matthew makes that particular sound, he means “frown”.

This is the point at which the person who has bought a package from a high-street shop might give up. Do the correction routine right and the system will learn the speech of someone who seems virtually unintelligible to others. Do it wrong and you get deeper and deeper into trouble.

“You can quickly corrupt your own voice file and make the system useless,” says Robin Korwin. So teachers need to be very confident with the software in advance. Uncertainty courts frustration and failure - disastrous for a child who has probably had quite enough experience of failure.

Robin Corwin says, “I always say that the child has to see success in the first session - words on paper, that they print out and take away, saying ‘I’ve done this’.”

There is growing realisation that the system can actually address literacy problems and not just help to work round them. Robin Corwin points out that “The child thinks the word, says it, sees it and hears it, and won’t move on until it is correct. The computer reinforces good practice.”

Stephen, also 12, has a long-standing problem with words. With voice recognition, he says: “I can tell the difference between the various ‘theirs’ now.” Say “their” into the microphone and the system will show you “they’re “there” and “their”. Having to choose the right one helps the child to learn.

But the biggest benefit of voice-recognition software is the boost it gives to children’s confidence and self-esteem. Voice recognition will surely change the lives of thousands of people with special needs. Teachers cannot ignore it. Even if schools are slow to take it up, it will soon be in pupils’ homes, supplied free with computers bought in the high street.

CHOOSING A SYSTEM

The early voice-recognition systems worked with “discrete speech” - you had to put tiny gaps in as you dictated. The most recent systems work with “continuous speech”. This is great for business and commercial users. For special needs applications, though, it may be a disaster if the development of continuous speech pushes discrete systems off the market. The hope is that a discrete system such as Dragon Dictate can be kept alive for education uses. A great deal of lobbying is going on.

You may need a new sound card. The software will state minimum system requirements, but as so often, the more power and memory, the better it will work. Laptops present more problems for voice recognition use than desktop computers. If you want voice recognition on a laptop, take advice from the people listed below. The quality of the microphone is crucial. It has to be carefully and correctly positioned, too.

Deployment of adult support in schools may need to be reviewed. All children need close support in the early stages. Some will need it for much longer. Teachers and other adults must be much more familiar with voice recognition than with most other software. They have to be able to solve the pupil’s problems quickly and confidently before frustration sets in.

There seems little likelihood at the moment that children will be using voice recognition in class. A Devon study found that a boy using a laptop in his science lesson, soon proved disruptive. “It takes a strong character to be able to dictate in a class situation,” the report commented. Background noise is an obvious problem, too.

There may be ethical issues. This is an expensive resource, and there are questions about who should have access to it. An adviser said, “Perhaps there really is a genuine equal opportunity issue here, and voice recognition should be available to all children.”

Do not buy this technology before taking lots of advice and, preferably, going to see it at work. It is easy to make expensive mistakes. Users like Robin Corwin constantly have phone calls from people who have run into trouble with systems they have bought off the shelf.

Prices vary from about pound;35 for cheap systems from big high street suppliers to up to about pound;150 for a well-tested system such as IBM Via Voice Executive, or Dragon Naturally Speaking Preferred (both continuous speech systems). Dragon Dictate Classic, at about pound;100 and IBM VoiceType Simply Speaking at about pound;50, are the discrete speech systems most used in education.

HELP AND ADVICE

The Devon VR Report: a study of the use of voice recognition with secondary dyslexic pupils. Readable and informative. From Sally Skinner, STEPS, 97 Heavitree Road, Exeter, EX1 2NE. Tel: 01392 276348. Cheque for pound;3.50 made out to Devon County Council.

Robin and Deana Corwin are always willing to give advice. They can also supply systems through “SayIT” Ltd, at Bilston College. Tel: 01902 821317 The British Educational Communications and Technology Agency is investigating the possibilities of voice recognition for children with special needs. Contact Mick Thomas, BECTA, Milburn Hill Road, Science Park, Coventry. CV4 7JJ: or www.becta.org.uk Iansyst specialises in the supply of dyslexia software (not just voice recognition) and complete systems. Its website has pages of really helpful information, with comparisons between systems, prices, and up-to-date news. Founder Ian Litterick is a mine of information. Website: www.dyslexic.com; tel: 0800 141515 or 01223 420101