The "killer" application that could fully justify buying an expensive, powerful computer may be as plain as the nose on your face, or rather, the mouth: speech.
Voice recognition software allows users to dictate, rather than type,
documents, emails, and basic commands. It also provides a natural way for users to interact with their computers.
Though it has been touted as the epoch-making application for the PC industry by analysts and users alike, it has yet to catch on
in large numbers. Chip giant Intel and computer makers have high hopes, as there have
been few compelling mainstream productivity applications that might justify the purchase of a PC with the most powerful Pentium II processors.
To be sure, there are still a number of technical and logistical obstacles that need to be overcome before speech recognition as a desktop application takes off.
The industry's move toward integrating more functions--like graphics and
sound--onto a single chip in the interest of cutting costs has resulted in
many computers, even high-end systems, with sound cards incapable of inputting voice commands, said William Meisel president of TMA Associates of Tarzana, California.
"Part of the difficulty with speech recognition right now is that not all
the sound cards support speech input," Meisel added. "Sometimes even the higher-end PCs use integrated chips, and integrated chips don't have good
sound."
Moreover, convenient headset microphones that allow users to switch from telephone to dictation modes are still too expensive. "The other bottleneck is the microphone itself. The ones shipping with PCs are good
quality, but not necessarily easy to use," Meisel noted.
With 450-MHz Pentium II processor systems arriving in stores this summer and 500-MHz systems not far behind, Intel and profit-squeezed PC vendors need to convince corporate and individual customers why they shouldn't just buy
a $799 computer with a processor from Cyrix, for example, or a low-end chip from Intel. (Intel is an investor in CNET: The Computer Network.)
"There's always a constant struggle to make sure that there's demand for
the processing power that's being put out," said Danny Lam, director of
Fisher-Holstein, and a speech recognition analyst. "You've got to find a way to consume it."
The system requirements from suites like IBM's ViaVoice, Lernout & Hauspie's Voice Xpress, and Dragon Systems Naturally Speaking are
robust enough to make Intel and PC
vendors excited about the prospects of selling more of the fast, pricey
systems.
For this kind of software, a lowly 166-MHz processor doesn't cut it. To run these applications with any consistency or accuracy, the system should have
at least a Pentium II 300-MHz processor and a minimum of 96MB of memory,
according to Lam.
"The [voice recognition] performance goes up as you add memory and processor speed," noted Lam. "It eats CPU cycles and memory. It consumes the power that they're putting out. Consumers are asking, 'Why do
you need more than a Pentium MMX?' and we found a reason why."
Microsoft has suggested that speech
recognition will be a large part of future operating systems, which will be
a huge push for the technology, according to Meisel. "Microsoft has come out pretty strongly about the importance of speech," he said. "[They've said] that they're going to put speech into the operating system, and they've put a
significant research effort into natural language recognition."
Even before Microsoft incorporates speech into their operating system and other technical wrinkles are ironed out, speech
recognition applications already have a real value to disabled individuals, especially office workers with repetitive strain injuries. Additionally, many
companies are finding adding speech recognition for phone systems can be a
huge money saver.
"Seventy percent of computers sold go to businesses," noted Lam. "For a business user, speech recognition is a performance tool as the accuracy rate gets to be around 94 percent. We find that for the average person, it beats
typing."
Also, for users who can't type at all, speech recognition applications
allow employees to operate computers effectively. "I'm working now, and I
was not working before," said Lewis Wallace, a Dragon user who suffers from
carpal-tunnel syndrome. Wallace says that he improves his efficiency by
setting up voice "macros," which can type a series of text such as HTML
coding with a single word.
Despite the positive impact voice recognition has had on his professional
life, Wallace pointed out an obstacle to the software that may not be
overcome even with technical improvements: eavesdropping co-workers.
"Frankly, my biggest problem is shyness. I felt somewhat embarrassed. Every
word you 'write,' all the work, all the email, can be heard," he said. "I
used to work in a large room with lots of people. There was a little performance anxiety that I never expected...but I'm in an office by myself
now."