Copywriter, technical writer, translator (FR>EN, ES>EN, IT>EN), journalist

Voice dictation tools helpful, but still have kinks

originally published in The Lawyers Weekly

Ronald Morton, founding member of the Clinton, MS Morton Law Firm, PLLC, relates his post-meeting routine: “I pick up my digital dictating machine as soon as a client leaves. I dictate what we talked about and what the plan is. As soon as I plug the recorder into my computer, the file is saved.”

“I have the audio and I’ll also transcribe it using Dragon (Naturally Speaking),” Morton continued. “I still have to correct it – it has some errors in it. But it’s good enough for my purposes, memorializing what a meeting was about.”

Morton describes two of the most popular uses among lawyers for software tools that respond to the human voice: voice recognition that helps people more quickly transcribe what they say, and digital voice recording that replaces magnetic tape-based recording.

Productivity is, of course, the main attraction. But litigator Maxwell Kennerly discourages people from using traditional measures of productivity to evaluate these tools.

“If you simply ran a stopwatch and compared how long it took to dictate and correct a document versus simply typing it, voice recognition doesn’t seem much faster and, indeed, is sometimes slower,” wrote Kennerly of the Philadelphia, PA-based Beasley Firm LLC in an email. “The critical difference is fatigue. After I type a document, I usually feel tired and unwilling to move on to my next task. Voice recognition software dramatically reduces that fatigue.”

“The difference is often greatest at the end of the day; instead of leaving your office with pain in your hands, wrists, and forearms, you leave feeling productive and ready to go back the next day.”

Voice recognition has successfully been integrated in other devices – voice dialling on mobile phones, for instance. But the current state of the technology has yet to equate that portrayed in Star Trek.

The first caveat: transcribing a meeting where multiple voices speak is beyond today’s technology. “It’s only going to recognize the voice that it’s trained for,” said Morton.

It also will not perfectly transcribe the “trainer’s” voice every time, though systems do improve over time “MacSpeech Dictate can reach a 96% or better accuracy rate right out of the box, with only a few minutes of training,” according to Sacramento, CA litigator Randy Singer.

“Training” software to recognize one’s voice usually starts with reading a prepared text to the system when first installed, then patiently and continuously correcting comprehension errors session by session.

Yet digital transcription may never fully replace typing. “Even with a 99 per cent accuracy rate, that means you have one error in every 100 words,” Morton said. “That’s still a lot of correcting to do.”

Kennerly added the following advice: “For improving performance in daily use, the question is not if you can get 100 per cent accuracy, because you usually will not,” he wrote, “but rather if you can adapt to checking and correcting the words on the page as they appear, which you normally do not do when typing.”

“For me, adaptation took about a week of frustration, after which I fell into a groove.”

Kennerly’s experience led him to the following conclusion: “Voice recognition is not for the computer illiterate,” he wrote. “You need some computer savvy to use it effectively, since your performance depends on your ability to quickly detect and correct errors, your ability to enable the more robust features of the program, and most importantly, your general comfort level with technology.”

Singer, who also heads the 7,000-member MacAttorney User Group, added disheartening news for Apple computer users: he claims that Dictate is the only voice recognition and transcription software product for the Mac that is being actively developed and supported.

Getting started with voice recognition calls for investments in both software and hardware. In software, all three lawyers mentioned Dragon Naturally Speaking. (Although he doesn’t use it, Singer wrote that MacSpeech is based on the Dragon speech recognition engine.)

For hardware, a separate microphone with noise-cancelling technology improves the performance of the system. That performance also rises in a quiet environment.

Certain microphones plug directly into a computer’s sound card, so better cards can help. Extra RAM in a must, since voice technology is, in Morton’s view, “a pretty intensive memory hog.”

Separate digital voice recorders let lawyers dictate away from their computers. Again, noise cancellation is indispensable in places like cars, where wind noise and air conditioning could make dictation unintelligible.

As an alternative to recorders, software designed for smartphones like the BlackBerry and iPhone let lawyers both dictate and email the resulting sound file to the computer where correction takes place.

Many lawyers will continue to send such files to transcriptionists, but Morton prefers to take them out of the loop. He gets his notes faster and “secretarial help can spend their time doing things that are more productive, more profitable for the firm,” he said.

To download a PDF of the published version, click voice_dictation_software.