Talking Computers Are Not Just For Games Anymore
This is a recurring vision: I'm driving down the highway with important work to do. I can't look at the laptop computer sitting on the front seat, so instead I listen to it read the day's important correspondence and background articles for the story I'm researching.
Wrong turn? No need to look at a map. The machine reads back the directions.
The technology to do this is almost ready to hit the mainstream. Our computers are about to become chatterboxes.
Already, computer games use recorded voices. And on-line services, such as CompuServe or America Online, use spoken greetings and alerts. What we want is a computer that recognizes words and says them out loud. This is called text-to-speech technology. And one of the companies aggressively pursuing it is Apple Computer.
When I got my Power Macintosh earlier this year, I was impressed by a little program called PlainTalk. You could type in text, pick a voice, and then have it read it back. Some of the voices were purposely weird. "Fred" was relatively normal, although he read English with a Swedish-like lilt.
(The program was an immediate hit with my wife, Gretchen, who normally abhors computers. She loved typing zany poems into the Mac and having "Fred" read them back to her.)
For the visually impaired, text-to-speech technology is crucial. Xerox, for example, is selling a stand-alone machine called the Reading Edge Express. About the size of a small copy machine, it reads books out loud. Users take a book or an article, scan it into the Reading Edge, then click a button to have the machine read it back to them.
I tried it out on two books: a 1950s expose of blacks in America and one of Gretchen's quilting books. I was impressed how well the machine deciphered fractions and handled proper names.
There were a few blunders. The Reading Edge tried to read TNNA, the acronym of an organization, as "Tinna," and applique came out "Ap-Leek." But these are minor criticisms for an impressive, well-integrated machine.
Talking computers are about to become more useful for mainstream users. For example: Have you ever typed a column of numbers into a spreadsheet and double-checked your work? It's awkward to read back-and-forth between the screen and the original piece of paper. Text-to-speech makes this a breeze. The computer reads the numbers on-screen while you read the original.
A few software companies already offer such programs, says Kim Silverman, Apple's principal research scientist for speech synthesis. Apple plans to incorporate such technologies in future versions of its operating system. The company is also working on computer-telephone applications so that a user on the road could call into her computer and have it read back her voice-mail.
Xerox, meanwhile, is working to create smarter readers. "As well as these machines can read a novel and magazines, there are pieces of paper that blind people encounter in their daily lives that they can't navigate through," says Philip Minasian, director of engineering and operations for Xerox Imaging System's adaptive technology products. After all, we don't need a computer to read back all the text on a telephone bill, just the amount due.
Similarly, many users are going to demand that their computers sound like real people. Apple's latest version of text-to-speech technology, renamed MacinTalk Pro, improves on Swedish-sounding "Fred" with new voices, such as "Victoria" and "Agnes."
"By the year 2000, we will have synthetic speech that, within the constraints of certain applications, will be virtually indistinguishable from human speech," says Apple's Mr. Silverman.
The silent era of desktop computing is about to end.
*Send comments via Internet (firstname.lastname@example.org) or to CompuServe (70541,3654) or America Online (LBELSIE).