Software That Takes Your Word for It
Like a lot of PC users, I’ve been known to talk to my computer, usually in expletives. Until recently, it never understood me.
But now, my PC can understand nearly every word I say, thanks to Dragon Systems’ Naturally Speaking. This remarkable software captures your words as you speak in a normal voice and at a normal pace. IBM’s ViaVoice, scheduled to ship late this summer, offers similar capabilities.
Both programs come with a noise-canceling headset microphone that plugs into your sound card and works better than most other PC microphones.
There’s nothing new about dictation software. Dragon, IBM and Kurzweil Applied Intelligence have long offered programs that allow users to dictate to their PC.
The trouble with these programs is that you must speak extremely slowly--one word at a time, with pauses between each word. This is known as “discrete speech recognition.” It works, but it is so slow and painstaking that it isn’t worth it, except for very slow typists or people who can’t use their hands because of an injury or disability, or because their hands are otherwise occupied.
With this new breed of products, dictation is now a real alternative for anyone who thinks it would be more productive than using a keyboard. These two programs work only with Windows 95; Mac users who wish to talk to their machine can use Dragon’s PowerSecretary, but they have to speak slowly because the program cannot understand continuous speech.
Naturally Speaking is able to keep up with a normal speaking pace. The company claims that users can achieve speeds of 100 words per minute, which is about what I experienced.
Before you begin using the program, you must train it to recognize your voice. The process is pretty easy, but it does take about 25 minutes while you read into the microphone so that the program can compare a sample of your voice against known text.
To ease the boredom, you get to choose a selection from “Arthur C. Clarke’s 3001: The Final Odyssey,” “Dave Barry in Cyberspace” or something really exciting like an employment request letter or a household management agreement.
The program requires you to do all your dictating inside a special text editor that comes with software. You can use either the keyboard or your voice to select text, make corrections and move about the editor. When you’re done, you can copy it to the Windows clipboard and paste it in a word processor, e-mail program or any other Windows application--all without touching the keyboard or mouse.
The program is far from perfect. When dictating this article, the software made more than 20 mistakes, but they were easy to correct. If you wish to correct a word with your voice, you tell the program to select it and either repeat the word or, say “spell it” and then spell it out loud. When you correct words or introduce new words, the program remembers what was done and is less likely to repeat that mistake.
Ironically, my biggest complaint about using the program isn’t its mistakes, but my own. I’m not used to dictating, so I found myself making mistakes in wording that I wouldn’t make when I type. People accustomed to dictating to a stenographer will probably feel at home, though a good stenographer is more forgiving.
My most serious mistake was when I decided to use my voice to copy the entire text to the clipboard so I could paste it in my word processing program.
I got off to a good start with “select document.” My next task was to copy it to the clipboard, but instead of saying “copy all to clipboard,” I said “copy to clipboard,” which the program interpreted as text. Like all Windows programs, it replaced everything I had highlighted with the new text, effectively wiping out my document.
I tried to undo that action by saying “undo,” but I should have said “undo last action” or “undo that.” Fortunately, I had saved it to disk. I now know to use the more familiar keyboard commands, rather than my voice, when working with large amounts of selected text.
Another problem occurred when my wife walked into the room and started talking to me while I was dictating. Naturally, I talked back, and the program made a vain but unsuccessful attempt to transcribe my half of the conversation. I should have turned off the microphone by clicking its icon or saying, “go to sleep.”
The program has some nice help features. Any time you don’t know what to do, you can say, “what can I say,” and it brings up a help box that you can navigate with the keyboard or your voice. There is also a multimedia tutorial, which gives you a good introduction to using the program.
IBM’s ViaVoice, scheduled to be released by the end of August, also allows you to dictate at a natural pace. I tested a pre-release version of the program, which, like Naturally Speaking, made plenty of errors. They were easy to correct with the keyboard, but there is no way to correct them by using your voice.
In addition to its own editor and input window, the program can be configured to work inside of Microsoft Word. However, installing the Word option adds a delay every time you start Word, which can be annoying when you don’t plan to use the dictation feature. Fortunately, it can be quickly uninstalled.
Later this year, IBM will release ViaVoice Gold, which will include correction features and the ability to control other programs by voice. These features are currently available in IBM’s Simply Speaking Gold, a $99 discrete voice-recognition program.
I’ve been typing since elementary school and have no plans to adopt either of these programs. However, I’m pleased to see progress being made in this field. To some, these programs will be a godsend. Long term, I expect the technology to improve dramatically.
With improvements in hardware and software, voice recognition will eventually revolutionize hand-held personal organizers, which now depend on tiny keyboards or imprecise handwriting recognition. Ultimately, it will work with pagers, phones, e-mail terminals and other devices and become a regular part of our personal and business lives.
Don’t even dream of using either of these programs unless you have at least a 133-megahertz Pentium and 32 megabytes of memory. ViaVoice requires a 150 MHz Intel Pentium with MMX. ViaVoice is priced at $199 and Naturally Speaking is about $299. However, both prices are likely to drop as competition heats up.
*
Lawrence J. Magid can be reached via e-mail at magid@latimes.com. His World Wide Web page is at https://www.larrysworld.com
