Software Worthy of Recognition
These days, it’s not just “Star Trek” characters who can talk to computers. Speech-recognition programs have evolved to allow you to dictate text at a natural pace and issue commands by voice--without using your keyboard.
Until recently, PC speech-recognition programs required you to speak very slowly. But the software has come a long way in the last year.
Two products, Dragon System’s NaturallySpeaking and IBM’s ViaVoice, allow you to dictate letters and other documents to your PC in a natural voice at conversational speed. Both programs cost about $100, come on CD-ROM and require a Pentium-class machine running Windows 95. IBM’s $149 ViaVoice Gold also allows you to issue commands to your PC and dictate e-mail and other documents directly into PC applications.
Lernout & Hauspie (https://www.lhs.com or [781] 203-5000) recently released Kurzweil Voice Commands for Microsoft Word, a natural language “command and control product” that allows you to issue editing and formatting commands to Word. Unlike the IBM and Dragon products, however, it does not let you dictate letters. The company said that by April it plans to release L&H; Voice Xpress, a continuous dictation and command program that lets you dictate documents and control your PC.
I came away impressed after testing both NaturallySpeaking and ViaVoice. They were able to keep up with my dictation and were reasonably accurate, though both did make mistakes. The programs require that you spend time training the software to recognize your voice by reading aloud into your PC’s microphone for half an hour or more.
For some people, these programs are a godsend. They can be a genuine boon to people who don’t know how to type or who type slowly, or for people with a disability that makes typing difficult or impossible. The programs can also help prevent carpal tunnel syndrome and other repetitive-stress injuries. Still, they’re not a cure-all.
I generally prefer using the keyboard when I’m sitting in front of my PC. But not everyone carries a laptop, and there are times--such as when I’m driving--when typing would be impossible or impractical.
Speech Machines (https://www.speech machines.com [650] 568-1500) has a solution for people who want to dictate memos and other documents when their PC is nowhere in sight. The British company, which has offices in Redwood City, Calif., doesn’t sell software but has just launched its Cybertranscriber voice-recognition service, which lets you dictate a document from a telephone or a hand-held digital recorder, then sends you the document by e-mail. Though companies or even freelance stenographers who take dictation by phone or audiotape aren’t unusual, Speech Machines stands out because it uses a combination of speech-recognition software and human proofreaders to deliver documents.
The system is easy to use. After you establish an account with the company, you dial a toll-free number, punch in your account number and secret access code and start dictating. Like ViaVoice and NaturallySpeaking, you need to tell the system when to punctuate and when you’ve come to a new paragraph, but you don’t have to spend any time training the system.
When you’re done dictating, you press the pound key on the phone and hang up. A few minutes later, the company sends you an e-mail message verifying that your job is being transcribed. The entire process take about three hours, although the company guarantees it will deliver the final document by e-mail by the next business day. You also can request the system to deliver the document as a standard e-mail message or as a Microsoft Word or other word-processing file.
Calls to Speech Machines are answered by a computer in Great Malvern, about 60 miles north of London. The company’s proprietary voice-recognition software does the best job it can to recognize your voice and turn it into text. It is then forwarded to a team of proofreaders on the Isle of Bute, about 15 miles off the coast of Scotland, who use software designed to detect areas of likely error. A proofreader corrects any errors and e-mails the document to the customer.
The proofreading process doesn’t guarantee an error-free document, but I was amazed to find only a couple of errors in several documents I dictated. In fact, it seemed to do a better job than PC voice-recognition software, and even did a good job with a memo I dictated from my cellular phone while driving down the freeway.
But it’s not cheap. The company charges a subscription fee of $9.95 a month, which entitles users to $14 worth of transcription. After that, the cost per double-spaced page is $3.50. There are discounts for volume users, but the charges add up quickly if you dictate a lot of documents.
Those who use dictation on a regular basis will find PC voice-recognition software a better value. In addition, PC-based software gives you the document immediately. Some users might be uncomfortable about possible security risks involved in having their documents transcribed by another company and accessible to human proofreaders. Speech Machines Chief Executive Mike von Grey said the company’s employees sign confidentiality agreements.
If you don’t want to dictate by phone, you can also use portable digital voice recorders from Voice It Worldwide (https://www.voiceit.com or [800] 478-6423), which start at about $200. These devices have ports that let you transfer voice files to a PC so they can be sent to Speech Machines for recognition. If you’re not in a hurry, you might save money by using a traditional dictation service. Hired Hand Transcription service (https://www.hiredhand.com) of Phoenix, for example, transcribes audiotapes for $1.95 a page and posts the typed document on a special Web site from which the customer downloads it. The downside is that you have to mail the tape to the company and wait up to 14 days for your document.
While I find current speech-recognition programs useful, I’m more intrigued by the long-term possibilities. Speech Machines, for example could use its technology to provide a voice e-mail service that makes life easier for travelers by letting them hear and dictate e-mail via phone.
Ultimately, I expect to see voice-recognition software that is able to run on future generations of hand-held personal organizers, such as the Palm Pilot. It will be awhile before such devices are powerful enough to recognize continuous speech, but we’re already seeing chips that allow you to use your voice to control hand-held devices. DSP Group of Santa Clara, Calif., (https://www.dspg.com) recently announced the availability of its True Hands Free Voice Recognition Chip, targeted at phone, cellular phone and auto makers, which would allow you to dial a phone or retrieve your voicemail without using your hands.
Web site links to all products mentioned in this article can be found at https://www.larrysworld.com/voice.htm
*
Larry Magid can be reached at magid@latimes.com. His Web page is at https://www.larrysworld.com or keyword LarryMagid on AOL.
