Something to Talk About
Users of next-generation personal digital assistants may find themselves hearing voices
By JEFF CHU
Look who's talking. It may soon be the Palm in your hand. Some personal digital assistants (PDAs) require users to write hieroglyphic-like shorthand onscreen with a stylus. Others reduce users to two-fingered typists hunting and pecking at elfin keyboards.
 |
Cyrus Deboo for TIME
| Neither interface makes much sense to Gary Saxer, company evangelist for speech-technology company Lernout & Hauspie. Speech, he says, is "magic. It's the natural choice."
Natural it may be, but, until recently, practical it was not. Pocket-sized computers simply haven't had enough processing power or memory to handle voice recognition software and data. But researchers at L&H and IBM, two leaders in the field, recently unveiled prototypes of speech-enabled PDAs.
The new devices allow users to perform tasks without lifting a finger. Say you need to look up Uncle Fritz's postal code. Instead of searching manually, you can say, "Open address book, find Uncle Fritz, find postal code." The PDA will read you the relevant details. While these handhelds aren't designed for full-scale web surfing, they are designed for network connectivity, and a range of online functions — from dictating e-mails to ordering pizzas — will also be possible.
L&H's talking PDA is called NAK, short for "nakulu," the Hawaiian word for echo. NAK, which looks like a hybrid of a PDA and a mobile phone, works its verbal magic using scaled-down versions of existing software — Voice Express for speech-to-text functions and Real Speak for text-to-speech — coupled with an Intel StrongArm II processor designed for handheld devices. Like other speech-to-text programs, Voice Express requires some "training" — about six minutes' worth — to understand new voices and accents. Most users need a couple of days of training themselves before they master the pausing that alerts NAK to forthcoming commands. As for Real Speak, it has an unlimited vocabulary and a velvety almost-human digitized voice.
IBM's PDA add-on, dubbed the PVA (Personal Voice Assistant), is a jacket that can be popped onto the back of an existing Palm-style device. The jacket houses a microphone, speakers, and the Embedded ViaVoice technology, which enables users to perform the PDA's basic functions with voice commands. A database storing 500-word sets of commonly-used words and commands also make the PVA speaker-independent. It can understand basic instructions from almost anyone, without training, whether the voice is young or old, high-pitched or low.
Of the two devices, NAK has greater versatility. Take message dictation, for example. L&H's technology transcribes a user's speech into text right on NAK, but the PVA doesn't have the horsepower to handle continuous conversion. The PVA user stores the dictation as an audio file, which is transcribed either by offloading it to a full-size computer equipped with ViaVoice or by sending it to IBM's WebSphere voice server.
Both prototypes exhibit the growing pains of gadgets in adolescence. If, for example, NAK doesn't understand what you're saying, you'll just have to repeat yourself until it does. The PVA doesn't always understand the user either, but if it can't figure out a command, it's trained to say, "I didn't understand that."
Refinements are on the way. IBM wants to add the capacity to dictate and send the kind of short messages sent to cell phones and pagers that have become so trendy in the European market. L&H would like to integrate ambient intelligence enabling NAK to make jumps in logic. If, for example, a London resident asks NAK, "How's the traffic?", NAK with ambient intelligence could look up the user's profile, note the hometown, and deliver traffic info for London.
Such gee-whiz features are too much for some in the industry. Jeff Hawkins, who helped create the original PalmPilot and now heads the team at Handspring, Palm's new competitor, has said that he doubts that users will choose speech as a preferred medium for data input. But William Meisel, president of speech industry consultancy TMA Associates, insists that speech is a "quick, natural method." He thinks consumers would buy voice-enabled PDAs, but only if learning to operate them "doesn't require a huge effort."
Voice-enabled PDAs will also face competition from the enhanced mobile phones on the way, but the companies are convinced they have a winning device. Tom Houy of IBM Voice Systems says, "When I hand these devices to people to play with, the biggest question is, 'When can I get one of these?'" Good question. Neither IBM nor L&H plans to make speech-enabled PDAs in-house. Both are seeking partnerships with manufacturers to integrate the technology into handhelds and market the enhanced devices. Consumers may see voice-enabled PDAs on store shelves as soon as early 2001.
But talk won't come cheap. IBM estimates the upgrade of an existing PDA will cost a couple of hundred dollars, a brand-new voice-enabled PDA at least double that. The price doesn't include recurring costs like airtime fees for accessing real-time information through wireless networks. Perhaps that seems a hefty price to pay to talk with an inanimate object. But think of it this way: at least these motormouths always come with "off" buttons.
TIME Digital table of contents
TIME Europe home
More stories from TIME Europe and related links
E-mail us at mail@timeatlantic.com
|
|
|
