Peter Sinclair: Speak up, and leave the rest to high-tech

"Peter Sinclair!" said Roger Cox, with the same sort of verbal flourish as you might pronounce the words "Open, Sesame!"

At once, I appeared onscreen. Or rather, everything about me did. There, in colour, were my name, address, date of birth and all sorts of other secrets, great and small, to which Roger is privy. For he is my doctor.

He is also a fellow geek, and the device which responded so obediently to his twanged command (well, he's Australian) is an object of extreme technolust called a Jornada 420, a Hewlett-Packard handheld incorporating, among many other marvels, voice-recognition that actually works.

With the correct cable plugged into the correct connection, he could have uttered my phone number and this obliging little genie would have rung me to remind me to keep taking the pills.

Another cable, another connection (for wireless connectivity awaits the next Jornada generation) and it would have disgorged either all or some of the day's data into his desktop PC.

Voice-recognition software was a darling of the 80s. Venture capital threw money at the idea in the belief it would produce a "killer app" for the blind and those who were intimidated by computers.

It never happened. Voice-recognition, with a few exceptions, rapidly moved to the back-burner and then off the stove entirely.

Today, though, with raging CPUs, the soaring popularity of mobiles and the development of the web itself, voice-recognition again promises to deliver the pot of gold those early investors hoped for.

The big name then was Dragon, but its software never really delivered - Roger's Tasmanian accent would have hopelessly confused it, for it was easily confused.

NaturallySpeaking, released in late 1997, garnered a slew of awards; but even so Dragon lost $22 million last year on revenues of $60 million.

Recently, it merged with the other leading voice-recognition house, Lernout and Hauspie, makers of RealText (text-to-speech) and Voice Xpress, which boast an accuracy of around 96 per cent, peaking at 98.

The resulting juggernaut will cover the field in areas to include smart-phones, handheld/mobile devices, automotive applications, voice-to-voice translation, English as a second language and multilingual text-retrieval systems - all potentially vital to the developing global market.

It will also pioneer development of the new science of "audio-mining," which sifts streamed material in much the same way a search-engine does text.

At the same time, L&H wolfed down Dictaphone Corp, and the simultaneous impact of these two acquisitions on its bottom line brought on a slight case of indigestion this quarter, a burp which is expected to be short-lived.

L&H won't have voice-recognition entirely to itself, though - the shares of newcomers in the field like Nuance Communications SpeechWorks and Tellme Networks are surging as they seek various niches in the technology.

Across the Tasman, VeCommerce is exploring its Antipodean potential. Already in place is a $A3.5 million system at Queensland's TAB which lets punters place bets over the phone with simple voice-commands.

Our own TAB, Mobil and Tranz Rail are also developing or extending services based on VeCommerce's expertise.

It's interesting, as I study the website of contender 21st Century Eloquence, to note that so far it seems to be gamblers and, yes, doctors who are having most fun with the technology.

"The Eloquent Physician and The Eloquent Orthopedist are Document Creation Assistants without equal.

"Using speech-recognition software, sophisticated database tracking, and extensive yet understandable templates, you can create the most 'Eloquent' of medical reports in less time than you currently spend dictating "

Now I know what to get Rog for his next birthday.

petersinclair@email.com

BOOKMARKS


MOST ALARMIST: VIP

The Voting Integrity Project has a lawsuit pending against America's first binding online public election, the Arizona Democratic Presidential Primary of last March, on grounds it violated the Voting Rights Act of 1965. Proponents of a brave new e-world may need to read this condemnation of online voting as being so insecure it is conceivable that unfriendly forces could rig another country's elections.

Advisory: richly paranoid.

MOST BLOODSTAINED: Ancient Korean History

Now that the Cold War is beginning to thaw, it might be an idea to examine the origins of a complex and turbulent country with which we plan to re-establish formal relations. This page will take you from Korea's mythic beginnings in 2333 BC to the death of King T'aejong of Silla, who unified the peninsula in AD 661.

Advisory: gore and archaeology.

MOST OPEN-SOURCE: Vorbis

In the excitement of Napster, many people overlook the fact that MP3 isn't, in itself, free - it's a proprietary format which annually makes millions for its developers, a private German company called Fraunhofer-Gesellschaft. Vorbis is billed as the new open-source alternative. There are no royalties and the reference software is free.

Advisory: listen and judge for yourself.

Links


Jornada 420

Hewlett Packard

Dragon

Lernout and Hauspie

Dictaphone

Nuance Communications

Speechworks

Tellme Networks

VeCommerce

21st Century Eloquence

Voting Integrity Project

Fraunhofer-Gesellschaft

Vorbis