Dragon NaturallySpeaking Preferred Edition

by Michael Hotek, mhotek@swynk.com
Source: swynk.com

I have been following voice recognition software ever since it first hit the market.   The early entries into this market were prohibitive expensive, lacked many features, and didn't work very well.  These early products were useful for a tiny niche market and were geared more toward accessibility.  The speech recognition products that are beginning to filter into the market are finally ready for true business application.

It's a little premature to go to the picture of everyone in the office talking to their computers, but it does have advantages for certain people.   People who do a lot of writing are prime candidates for this technology.  In fact Dragon NaturallySpeaking was used to construct this review in about 20 minutes and has been used to put together much of the content in my section.  I have also begun participating in writing books.   I have yet to write one completely myself, but I have produced numerous chapters for two books with a third one getting ready to start up and a fourth on in the planning stages.  Dragon NaturallySpeaking has enabled me to produce hundreds of pages of content in fraction of the time it would normally have taken.

For this review, I'll dispense with the screen shots since there really is very little to see.  It comes with a stripped down word processor.   This is your best bet for performance purposes.  I have used the NaturalWord feature which integrates NaturallySpeaking into MS Word and allows you to dictate directly into a document.   Even though it worked well, the performance was very sluggish on a system with 128MB RAM.  My suggestion is to have both Word and NaturallySpeaking's word processor both open.  You then dictate into NaturallySpeaking and periodically copy and paste the text into Word.

NaturallySpeaking is also hands off and does require use of arcane commands and speech.   You have full control over menus, dialogs, navigation, and other features completely by voice.  Saying: Click OK will cause the OK button to be clicked.   Saying: Go to end of paragraph causes the pointer to move to the end of the paragraph.  Moving text from NaturallySpeaking to Word is very simple and intuitive.   You simply say the following: Copy all to clipboard, switch to previous window, paste that, switch to NaturallySpeaking.  Notice that the command was switch to previous window.  This means that any application that is open can be activated and have text pasted into it. 

In addition to the navigation features, this is a very solid product.  The whole point of speech recognition is accuracy.  This is accomplished through initial and ongoing training.  The first phase of training is simply to test your sound system, microphone position, and volume.  You are then instructed to speak a small series of sentences after which a very rough adjustment is made to the speech files.  The second phase then commences.  During this phase you read paragraphs from a variety of selections in order to calibrate NaturallySpeaking to your voice.  It will measure approximately 30 minutes of speech and then make adjustments.  Once the second phase of training is completed, you are ready to begin.   Before starting into dictate, it is recommended that you take advantage of the customization feature.  During this stage, you select a variety of documents on your system that represent the types of documents you will normally dictate.   NaturallySpeaking then scans the documents looking for words that are not in the dictionary.  You then go through some additional training where you add these words to the active vocabulary and train NaturallySpeaking to recognize them.  Don't be inhibited here.  Quite a few of the additions that I have made are not words at all, but functions.   NaturallySpeaking will now recognize SQL, xp_cmdshell, DBCC, rowfocuschanged, and a host of others as easily as it recognizes the, I, in, and is.

NaturallySpeaking is equipped with a 230,000 word vocabulary, 42,000 active, with the ability to add up to another 30,000 words to the dictionary.

During your dictation, it is inevitable that NaturallySpeaking will not recognize some words or phrases you dictate.  When this occurs, you simply tell NaturallySpeaking to select the incorrectly recognized text and go into a training mode where you train it to recognize the word or phrase.  Over time the occurrences of misrecognition will gradually decrease.  Each time you close a dictation session, you are prompted to save the speech files.  This will save any new words that have been trained and will also make minor adjustments to the speech files.

It is supposed to be able to recognize up to 160 words per minute.   I have not achieved that speed yet, but I am approaching those levels.  As you use NaturallySpeaking more often, you will notice a marked increase in performance as the speech files get more tuned to your voice.  When I first started working with it, I was achieving about 90 words per minute at approximately 95% accuracy on the technical information I was dictating.  This is still a far cry from the approximately 40 words per minute that I can type.  Over time it has gradually gotten better and now can recognize almost 130 words per minute at nearly a 99% recognition rate.  From these figures, you can get an idea of why I really like this piece of software.  What used to take me hours to produce can now be done in minutes.  This allows me to get more information out to you on a more timely basis.  It also greatly adds to the creative process.  There is no way I can type a fast as I think.  With the dictation speed of NaturallySpeaking, I can more logically follow a series of thoughts and get them on to paper for others to use.

One new feature that is supposed to be available in the most recent edition is dictaphone support.  With this you can get an incredibly powerful system.   You simply purchase one of the approved hand held dictation devices which are industry standard and not proprietary.  Then as you go throughout the day, travel to work, sit on the beach and relax, or go for your workout, you can dictate documents into your recorder.  When you get home, simply plug in the recorder, hit play, and watch as Dragon NaturallySpeaking transcribes your documents.  Talk about being able to squeeze every useful minute out of your day!

Dragon NaturallySpeaking is produced by Dragon Systems.  The Preferred Edition sells for $229.00.