They may find keyboards a misery, but there is one aspect of office practice with which they are familiar - dictation. That is why the largest crowds at a legal technology exhibition are usually around the stands demonstrating speech recognition software- systems that allow you to operate a computer by talking to it. But just how viable are these systems and, bearing in mind that the first few made their debut nearly three years ago, why aren't more lawyers using them?
At the heart of every speech recognition system is a mathematical software model, known as an engine, that converts analogue speech into a digital code that can be understood by a computer. It "hears" the sound through the microphone into which the operator speaks.
The engine is not so important, however, as its integration with the applications being run on the computer. Typically these fall into two categories: speech-to-text, where the spoken word is converted into word-processing, and command-and-control, wher e speech is used to tell the computer what to do, such as open a file or print a document.
There are four main speech recognition engines available in the UK, from IBM, Kurzweil, Philips and DragonDictate. The latter is the market leader, found in an increasing range of software applications aimed at lawyers. The Kolvox LawTalk system (based on DragonDictate), for example, will shortly be available as a way of controlling accounts and financial management programs for solicitors and barristers.
But it is the speech-to-text applications that attract most interest, as they appear to offer a seamless link to the way so many lawyers already work. Rather than preparing a document by dictating it on to tape for a secretary to transcribe and return for checking, it can be dictated into a word-processor and revised on screen.
Michael Pettman, a solicitor in Knightsbridge, was so impressed with a DragonDictate system he saw in the early 1990s that he bought the rights to the product and set up his own company, ASA Voicewriter, to sell the concept of speech recognition technology to the legal community.
Instead of waiting for tapes to be transcribed, documents could be turned round almost immediately. The system could reduce a firm's administrative tail by improving the secretary-to-lawyer ratio to one secretary for every two or even three fee-earners. This allowed recruitment of additional fee-earners without having to employ corresponding numbers of support staff.
Alternatively, the systems can be used in conjunction with desktop publishing case management software, as Bill Meads, a partner with Coffin Mew & Clover in Portsmouth, found. He uses the Allvoice system (also based on DragonDictate). "If I didn't have the system the only way I could get through my caseload would be by employing an additional secretary," he says.
Yet despite the apparent attractions, speech recognition technology has failed to make the expected impact, with far too many firms buying one-off systems for pilot projects that are never followed through.
One problem is the cost. Prices have fallen to around a third of what they were three years ago - a complete hardware and software system costs between £2,250 and £3,000. But this is still considerably more than a conventional PC running word-processing
Another negative factor is that to avoid the system picking up background noises, the operator has to wear a headset microphone; this is difficult if the telephone rings, or you need to break off dictation to speak to someone in the room. But this is notinsurmountable and Philips says that its new system, which goes on sale in the spring, will address many of these ergonomic issues. What does remain a problem is the time needed to master the system, in particular getting it to recognise each user's speech patterns.
According to Kurt Lynn of Kolvox: "Voice technology has to deliver an almost immediate improvement in productivity so the effort put into mastering it is well rewarded." But how much effort?
The system has a vocabulary, and you teach it how you pronounce each word. This has to be done in a consistent and natural fashion so that the next time you say the word the system will recognise it. In many cases the system gets it right first time, butsome words have to be repeated time and time again. Indeed it is nearly always quicker to type in the word than wait for the system to guess it - either that, or resort to an entirely unnatural Shakespearean tone.
Along with mastering speech recognition, the user also needs to be familiar with the system's applications, which suggests that if you are going to have to learn to use a program such as WordPerfect anyway, why bother with the speech element?
Some of the promotional literature gives the impression that you can be talking to your computer in a couple of hours. After some prompting, however, suppliers reluctantly concede that you need 15 to 20 hours' training to reach an acceptable speed, and one user found it took him nearer 60 hours.
And when should this training take place - in the lawyer's own time at the end of a long day? Or during office hours, bearing in mind that even 15 hours' training equates to between £l,500 and £3,000 worth of lost billable time.
Speech recognition systems are making great strides and the day will come when voice input becomes as widely used as keyboards are now. But for the time being, the majority view among lawyers probably echoes that of Courtney Parks, the IT manager at Rowe& Maw: "It is too expensive, too slow, takes too long to train and is not accurate enough. Perhaps in the next couple of years ..."
The writer is the technology editor of the Law Society's `Gazette' and technology journalist of the year.Reuse content