Mark Bagshaw, international marketing manager for IBM Australia, has increased his typing speed from about 20 words a minute to nearly 150 in three months — by using his company's latest voice recognition package, ViaVoice 98.
Bagshaw, who is quadraplegic, can type "as efficiently as most of the people I work with" using one finger but decided to give speech recognition a go when he started to feel mild OOS (occupational overuse syndrome) symptoms in his elbow.
"That's most of the movement I've got," he says, "so I have to look after it." Changing to talking from typing has been a pleasant surprise, he says. "It's made a huge difference to me and I've actually been surprised at how well it works." Never having used a speech recognition package before, he says he expected "to have to ... speak ... one ... word ... at ... a ... time, but it really isn't like that. If it had been, I think I'd have lost my train of thought but I talk at a normal speed, in a normal tone of voice and it picks up most of it. It's getting better all the time as I train it," he says.
The "standard" voice built into the system wasn't very useful, he says, because its British pronunciation is too different to his mild Australian accent. He says an initial 15 or 20 minutes was enough to get started, and then the full training — adding phrases as they appear on the screen until ViaVoice learns your personal voice pattern — took about two hours.
ViaVoice 98 lets the user scroll through menus, choose options and do most things possible with a keyboard and mouse. It has a base vocabulary of 56,000 words and space to add 64,000 more. Industry-specific vocabularies, including one for medical English terminology, are being developed.
Taking things a step further, Speech Recognition Systems has launched a handheld recorder designed for speech recognition transcription.
Dragon Naturally--Speaking Mobile has been designed for people who use dict-a-phones regularly. They can create, edit and format documents while speaking into a handheld recorder. The documents are then downloaded to a PC where they are transcribed, at a faster speed than they were originally recorded.
Will voice recognition systems eventually take over from keyboards? The only drawback Bagshaw can see is the noise levels that talking to computers would create, "though it's unlikely to be any worse than a room full of people on the telephone".