Voice Recognition Improves, but Keep Your Mouse Handy
About eight years ago I was on a radio program when someone asked whether we would have accurate and fast speech-recognition systems for our computers any time this century. I said no, because I believed it was a long way off. I was wrong.
Programmers at Dragon Systems and IBM have done an incredible job developing software that is able to understand the spoken word. And, unlike previous generations of speech-recognition software, you don’t have to speak slowly for the computer to comprehend what you’re saying. IBM’s ViaVoice and Dragon Systems’ NaturallySpeaking are designed for “continuous speech” recognition at up to 160 words a minute.
Both companies offer a variety of versions ranging in price from about $50 to about $200. You can find details at https://www.dragonsystems.com and https://www.software.ibm.com/speech.
While the newest versions of these programs are quite good, they are not necessarily going to improve your productivity. In fact, after learning to use both programs, I’ve removed them from my machine because I’ve decided that I prefer the keyboard. But just because I don’t like dictating to my PC doesn’t mean that it might not be right for you.
The first thing you need to know about continuous-speech-recognition software is that you can’t use it right out of the box. As with any sophisticated piece of software, you have some learning to do. But unlike most programs, the software itself has a lot to learn before it can do its job properly. Today’s continuous-speech-recognition programs must be trained to recognize the way you pronounce words. Training involves reading passages of text from the screen into a microphone for at least half an hour so that the software gets used to the way you speak. Even then, it will make mistakes when you start using it but, over time, the programs learn to understand how you say regular words and pick up special words and names that you use even though they’re not part of the vocabulary that ships with the software.
Both programs come with a headset that includes a noise canceling microphone and an earpiece. The headset has two plugs that connect to the microphone and speaker sockets of your PC’s sound card. The headsets are similar to ones used for hands-free telephone use. A $99 adapter from VXI ([800] 742-8588) lets you use it as both a PC and phone headset.
Dragon Systems pioneered continuous-speech recognition when it released the first version of NaturallySpeaking in the summer of 1997. At the time I was impressed but noted in my review that it made a fair amount of mistakes. It still makes some mistakes, but relatively few.
All in all, it’s finally ready for prime time. Anyone who wants to “type” by using voice instead of fingers can now do so with an acceptable level of accuracy. It’s also faster than before and easier to use. Version 3.0, released this fall, works from within Microsoft Word and other programs so you don’t have to copy your text from the software’s editor into your word-processing program.
IBM’s continuous-speech-recognition software has also been around since 1997, and it too has improved considerably. ViaVoice 98 Executive, which I used to write part of this article, isn’t quite as accurate as NaturallySpeaking, but I found it a bit easier to use. The program makes it easy not just to dictate text but also to control your PC and applications as you work. You can use it to launch programs and format your work. If you say “bold this,” for example, it will automatically put the subsequent text in bold type.
Although I can recommend either program for anyone who wants to dictate to a PC, I’m not convinced that these products are all that useful for most people. To begin with, they take a fair amount of patience to train, and, for those of us used to typing, it might be easier to stick with what already works.
Even after I was able to get pretty good accuracy, I often found myself reaching for the keyboard and mouse, especially when I wanted to enter small amounts of text.
What’s more, dictating is a different art form than writing at the keyboard. I find that I can write better when I type than when I speak. For reasons I don’t quite understand, I find that I get tired more quickly when I dictate my letters and articles.
Also, dictation software, despite its gains, still has a long way to go before the computers on our desks are as responsive to our voices as the ones used by the crew of “Star Trek.” You can’t just walk up to a PC and ask it to take a letter or perform a task. You must first put on a headset, run the software and make sure that you’re in the right mode for the machine to do whatever it is that you want it to do.
The software is a boon to those who can’t or don’t want to type. It’s great for people who have physical disabilities that prevent them from typing; it’s also a way to avoid repetitive stress injuries. I know several writers who are using programs like NaturallySpeaking and ViaVoice to prevent aggravating hand or wrist injuries.
Dictation is also a better way to communicate with a computer when typing is either impossible or inconvenient. Doctors, for example, can use dictation software while examining patients. Police officers can dictate reports from a crime scene or from their patrol cars.
Until recently you had to be at a PC to use these programs, but Dragon Systems now offers Mobile ($299). It comes with a portable digital recorder that lets you dictate on the go and then connect it to the PC’s serial port to upload the speech files, which are then recognized by the software. I found the mobile systems at first to be less accurate, but I was able to get better recognition by retraining the software under conditions that simulated dictating from a car or a noisy conference hall.
*
Lawrence J. Magid can be reached at larry.magid@latimes.com. His Web page is at https://www.larrysworld.com or keyword LarryMagid on AOL.
