Computers: Spelling out the advantages of a background in classics: Kevin Carey finds voice-activated systems have more scope than just helping dyslexics

Gone for good, I suspect, are the days when all three of us in my office had Latin A-level and diverted ourselves with etymological teasers. We are learning to take bad spelling for granted; decrying it in the young while hardly noticing how widespread it is among our peers.

For most of us, incorrect spelling is at worst an irritant and at best a source of modest mirth. But for the profoundly dyslexic, it is a serious handicap. There are hundreds of highly intelligent people who have a lot to say but cannot get it down on paper. For them, voice-operated wordprocessing is a dream come true.

Voice-operated computer systems have been steadily gaining ground while, symbiotically, falling in price. The first in the field were the Japanese, faced with the almost impossible task of using standard keyboards for expressing their pictographs. But it was not long before Dragon in the US saw the possibility of using them in business. Initially, systems were linked to word-processing packages, but now Dragon has linked its system to Windows, the PC's graphics-based, point-and- click operating system.

Aptech in Newcastle upon Tyne became interested in the Dragon Dictate system because its prime concern is computer technology applications for disabled people. Peter Kelloway, Aptech's managing director, immediately saw the possibility of helping dyslexic people.

The basic hardware configuration is two desk-top PCs linked together. The first of these handles the Windows package, while the second handles the Dragon Dictate system. The operator speaks steadily into a lapel microphone concentrating, in the first instance, on the Windows screen showing dictated text.

When the programme fails to interpret a word correctly he or she switches attention to the second screen, which shows a list of 10 optional words. If the correct word is there the user says 'Choose', for example, '3'. If the word is not there, the user says 'spell' and dictates the first letter of the word. The second screen changes its options, having taken account of the first letter. Again, if the correct word is not shown, the user chooses a number. If not, the second letter of the desired word is spelt in, and so on.

This might seem to defeat the object of the package, but two important factors need to be mentioned immediately. First, users build up a personal voice file as they dictate and every correction alters the package's sensitivity; it will work for somebody speaking BBC World Service English or with a rich regional accent.

Second, the package has a foreground dictionary of 25,000 words - words it immediately recognises from dictation - and can accommodate 5,000 more words personal to the user. It can also delve into a 'background' dictionary of 110,000 words when presenting its numbered options. Thus, when I started dictating, it got my village of Hurstpierpoint on the third letter spelt in.

The start-up process can be immediate, but the first few weeks will be slow going. You cannot put in a whole passage of purple prose, mistakes and all, then go back and correct because the package is designed to learn from its mistakes - you have to correct as you go. Every time an error in interpretation presents itself, you have to put it right. Over time, the interpretation error rate decreases and dictating becomes much smoother.

After a while you build up your own comprehensive voice file. Thus, if, for example, your office has the proverbial trio of an Englishman, an Irishman and a Scotsman, each will log on to their own voice file before starting dictation. The process is so sensitive it will distinguish between 'save' uttered in normal speech as part of a document, and 'save' uttered in staccato style as a prompt.

The advantages for the dyslexic are obvious; many do not know how to spell, but know when what is on the screen 'looks right'. For those who are more profoundly disadvantaged, a voice synthesiser output can be attached which will speak back what has been dictated, either a word at a time or spelling letter by letter.

The system has wider applications than helping dyslexics. Executives who do not like using a keyboard can dictate comfortably. Businesses, like law firms, that use a large number of standard paragraphs can build up 'macros' or short-cuts - so that dictating 'Standard Paragraph 1' might produce: 'With reference to your most recent letter, I am pleased to inform you that your analysis of the case is correct; Article II of the convention applies.'

People working in sanitised environments - or dirty environments - could use a microphone for dictation while looking at the computers through a window.

Building voice macros has another advantage; it allows people familiar with one word-processing package to use a totally different one without the problem of converting files from one format into another. 'Save document' will do the trick in WordStar or Wordperfect.

Operators are not, of course, confined to a completely voice-operated environment. Some people will want to dictate using the voice function, but may want to use the keyboard for prompts or for editing. Using voice commands to get the cursor round the screen will be a boon for somebody with a keyboard phobia or for somebody operating a totally strange word-processing package, but for many users it will be painfully slow.

The price of the basic components is high and while prices are coming down, there will be some who find them prohibitive. On the other hand, instead of employing a secretary with the proverbial Latin A-level, you might dictate letters yourself and merely pass files on to a less well-qualified person for tidying up, printing and despatching.

As long as these systems are few and far between, the flexibility will be limited, but as they become more widespread you will be able to carry your personal voice file from place to place. As long as there is somebody around in the office to 'tidy up', you will even be able to dictate memos down the telephone without feedback screens. Whether this will compensate for, or cause more, traffic jams on the M25 is a debate for another time and place.

DRAGON DICTATE

System: PC-compatible

Requirements:

Hardware: PC with 486DX or better processor

Software: Windows.

Publisher: Aptech.

Availability: Aptech 0661-860999

Price: List: Circuit board and software: from pounds 750 (exc VAT) for Dos to pounds 3,500 (exc VAT) for Windows depending on size of vocabulary.