Voice recognition – especially for emails

Voice recognition has progressed enormously from the early days when you had to spell everything using the phonetic alphabet, alpha, bravo, etc. It is now a very useful tool if used in the correct way.

My views are shaped by what I do as a sole practitioner in a small rural town offering the traditional Scottish “man of business” service, but principally, conveyancing and estate agency, agricultural law, partnerships, wills and executries, as well as fulfilling a fairly significant social function for those who have no other opportunity to air their personal problems.

What to use it for

If you are dictating everything that you create at the moment, voice recognition is going to add very little value as it is more time consuming than dictating onto a tape or solid-state device. Clearly, whether it is cost effective for you to use voice recognition for your major work depends upon the relative costs of yourself and the staff who will actually type the work. If you are charging £250 per hour and you are having your work typed in India at £5 an hour, there is little point in voice recognition.

I have two paralegals, two full-time and one part-time secretaries which will give you an idea of how little of my output is voice recognised. It is my view that voice recognition should be used only for unique work. I never want to do the same thing twice and if I am likely to say the same thing to another client in the future or produce the same document, I take the time to add it as a template and use it again when needed. Not only will it be correct and, hopefully, complete but I have saved myself a great deal of future time and angst.

So, what do I use it for? Principally emails. Most emails are non-standard and 50 per cent of mine are about office administration rather than client communication. However, for clients, the ability to provide instant feedback in between seeing clients is a great boon. I certainly couldn’t produce as much with one-fingered typing.

Purchasing a system

The two major providers of voice recognition software are IBM ViaVoice and Nuance Dragon Naturally Speaking. I have tried both over the years and am currently working with NaturallySpeaking Version 8 Professional. I recommend that you buy a system that includes a microphone and don’t buy cheap: you will regret it. That said, prices for standard software run from £20 to £160 for the mobile edition.

Incidentally, for those who work with solid-state recording devices, the software will automatically recognise anything you have dictated out of the office, when you come back. I haven’t much experience of that but I have recently purchased one and will experiment with it when it arrives.

Since my daily commute to the office is less than 100 m, I don’t foresee much use when travelling but it might be that I could dictate work in the office, have the software recognise it and only then give the partially-completed work to my staff for them to top and tail. Any thoughts and experiences from readers would be appreciated.

Learning to use it

The software learns to recognise your particular voice as a result of its mistakes and it is essential that you correct these as they arise or the software will reinforce the errors. The manufacturers claim 99 per cent accuracy but I’ve never experienced that. Nevertheless, I love it as it gets the business done.

Correcting is very easy. I merely say “correct” and then the word or phrase to be corrected, a box appears, you correct the word and that’s it. I don’t type the corrections, merely say them. I have mastered the phonetic alphabet which makes it easier but you can just say “r-u-b-b-i-s-h” which is quicker than typing. As a general rule, I avoid going near the keyboard. The software allows you to jump about your document by voice alone.

I recommend that you invest in a powerful computer with a significant amount of RAM. If you take too much notice of the minimum specification found on the software box, you will be will sitting waiting for ever for the words to appear on the screen. I have a Pentium 4, 3 GHz with 3 GB of RAM and, as I dictate this article, the words are appearing on my screen instantly.

Installation has always been very simple and the training, these days, only takes a few minutes. I work in my own room which is quiet (when the phone stops) and it might be more difficult for those who don’t have that advantage. The software is quite sensitive and is always listening. The fact that you can dictate virtually straight from the box would appear to indicate that it will recognise all sorts of accents.

I do have a problem in that whenever the phone rings I have to take the microphone headset off and put my telephone headset on. (I have a Panasonic DECT portable phone with a wired headset so that I can walk around the office and remain available.) I have been searching for a solution whereby I can use the same headset for the telephone, Skype and voice recognition. Any ideas will be gratefully accepted!

Bruce de Wert is principal of Caithness firm Georgesons.

Email bruce.de.wert@georgesons.co.uk.