Voice recognition in practice

I had first thought about getting voice recognition software as long ago as 1996, and while I had been impressed by the wizardry, I had not been impressed by the functionality.

Having kept an eye on the reviews since then, particularly those that appeared in PC Pro (in my experience it provides a balanced and sensible view of hardware and software), voice recognition software finally seemed to have become a functional and worthwhile product.

However, I would not have started using it were it not for the fact that, in mid 2007, I signed up for a two-year course which involved the submission of a number of online assignments, which I would be working on outside business hours and which I would need to submit outside hours. I therefore did not want to rely upon office support, even if I could obtain it.

I therefore seriously considered, for the first time, the use of voice recognition software to enable me quickly to draft my own documents without relying upon slow four-fingered typing. However, I also appreciated that this gave me the opportunity to experiment by using it in the office.

Choice of product

Consideration of the PC Pro reviews suggested that Dragon NaturallySpeaking would be the right product for me, so the question was how I should buy it. An email round the LawNet IT group came up with a reasonably local supplier who would supply, install, train and advise. I duly turned to Cameo-IT in Wantage for advice.

The first choice was whether I needed the more expensive Dragon NaturallySpeaking Professional with its additional special vocabulary. Cameo-IT did not push this and I decided I could do without it. This has proved to be a sensible choice since the standard software (Dragon NaturallySpeaking Preferred) can recognise words you often use, even if these are not common in English.

The second piece of sensible advice was to upgrade the memory of the computers I have at home and in the office. (For one licence fee I can use the software both at home and in the office.) I have an Acer Aspire 1800 in the office which runs Windows XP Professional Service Pack 3 on a Pentium 4 3.2GHz processor and 1GB of memory. The advice was simply to increase the memory to 2GB, which I did, and I also doubled the memory on a similar PC at home to 2 GB.

As a result I do not see any real delay in the transcription of the text into Word or any other application, even if I have five or six other applications running at the same time.

I should make it clear that I am running Dragon NaturallySpeaking version 9.51, and the newer version may be more slightly more memory hungry, although it is reputedly even more accurate. However, my satisfaction with the current version is shown by the fact I have not deemed it necessary to upgrade.

The cost of the software, together with hardware (new speech microphone), installation and training by Cameo IT, was £750 and well worth it. The cost was made up as follows:

Dragon NaturallySpeaking Preferred v 9 (with CD, manual, headset/microphone) £128
Philips SpeechMike Pro USB £189
Half day on-site training £295
Travel fee per visit £25
Total Cost including VAT £748

I understand that the most up-to-date version 10 retails at £148 excluding VAT. At this price, for those already using digital dictation and who therefore have some kind of digital speech microphone, it is probably worth purchasing Dragon NaturallySpeaking and installing it yourself as an experiment.

Installation and training

Installation was straightforward and there followed around two hours’ training on the software and how to use it effectively, which included training the software itself to my voice. This takes place simply by reading a selected text which appears on the screen. You then have the option for the software to scan one or more of your folders to enable it to pick up on the words you use and then to add them to its dictionary, although it has a very extensive dictionary of its own.

One of the benefits of having a third party supplier install the software and to train me was that they could take me through the typical commands – which are not complicated – and explain the set up. I am sure I could have worked it out myself, but it would have taken far longer to fiddle around and experiment.

Use and benefits

How have I used it since I got it 18 months ago? It has indeed proved to be very useful for the online course and I have dictated all my numerous 2,000 to 3,000-word assignments without any difficulty, far more swiftly than I could ever have typed and with an enormous degree of accuracy.

It is possible to dictate into almost any Windows-based application and although I typically use it in Word, I can dictate into Lotus Notes and into the text boxes in our office accounting system with a great degree of accuracy.

It is possible to use the software without the keyboard since you can dictate menu commands to bring up the menus and then choose from those. However, this is one of the areas to be wary of because if you simply say, for example, “file” as you dictate slowly, the software will go to file on the menu. However, if you say the word “file” within a long sentence that you dictate in one dictation, it will simply appear as a word in the text as you intended.

In fact this is one of the ways in which the accuracy of the software can be improved, by dictating a reasonable section of text in one go: seven or eight words or more, not simply one or two words at a time.

In addition, the software is trained with your voice against the ambient background. For this reason it is worth initially training it in your typical office environment. As it is used it will adapt to any slight changes. However, if you are likely to use it in more than one environment you can train it in different locations and have two different users, switching between them depending on location: for example, “James in the office” and “James at home”.

You do not have to dictate like a robot as you used to with the early versions, but enunciation is important and one of the benefits of this is that you are likely to be slightly better at dictating to your own secretary.

However, this is one of the things that you need to be cautious of, since it is not foolproof and if you do not enunciate your words clearly then it will type something that may not be what you said. It is therefore important either to watch the software as it types, or to read carefully what it has typed when you have finished the document.

The other thing that is easy to forget is to switch off the microphone. If you are interrupted during the course of a dictation and you do not switch off the microphone it will continue to attempt to transcribe, and will not only keep typing in the document that you are drafting, it will probably also start opening menu commands as it tries to interpret your remote voice.

One of the notable features of Dragon NaturallySpeaking is that it types far more quickly than any typist could and keeps up with a very reasonable dictation speed.

Its accuracy improves with use, because it learns both the words you use and the way you use them in each session. In this way it recently had difficulty with the name of a new client called Hermione, but after two or three uses it now spells this without any difficulty.

One of the impressive features of it is the fact that it considers the meaning and the context of what you are saying and therefore, for example, capitalises names automatically and distinguishes between “there” and “their”.

Dragon NaturallySpeaking has proved to be very useful by enabling me to do numerous tasks during the day when my secretary is engaged on something else important or is otherwise unavailable or outside office hours.

Inside office hours I am able to:

  • type my own urgent letters (usually by copying a previous one and amending that to save having to dictate name, address, reference and so on)
  • type quickly both short and long emails
  • make long amendments to letters, witness statements and other documents drafted by my secretary.

Out of office hours I am able to type and send out letters, emails, witness statements and any other documents at any time, which enables me to ensure that urgent matters are completed within a reasonable time even if they cannot be completed within the standard working day.

Dragon NaturallySpeaking has numerous features which I have not fully explored. I know that I could make more of the product: making it more accurate by training it more efficiently, and setting up simple macros, for example, to enable me to use my own templates.

The end of legal secretaries?

Cameo-IT told me that they had just finished an installation at the John Radcliffe Hospital as a result of which all the consultants at the hospital were all now using Dragon NaturallySpeaking and they had virtually no secretaries.

However, I do not believe that Dragon NaturallySpeaking will make legal secretaries redundant. It cannot intelligently work out, for example, that you in fact mean Defendant when you say Claimant, nor can it cope with hastily (and badly) enunciated dictation. It cannot print the letter, copy the enclosures, and put it all in an envelope and then file everything away.

It can, however, act as an extremely useful and efficient way to extend fee earners’ ability to generate their own work, and at this price it is worth the investment.

I have of course dictated this article on Dragon NaturallySpeaking and it correctly spelt, for example, LawNet, Wantage and John Radcliffe.

James Couzens is a Principal of Parrott & Coales LLP in Aylesbury, Buckinghamshire. He is head of the firm’s litigation department and is also in charge of the firm’s information technology.

Email james.couzens@parrottandcoalesllp.co.uk.