Speech recognition goes down the drain

Published Jan 13, 2010

Share

Regular readers will know that I have been using Dragon

NaturallySpeaking speech recognition software for a couple of years to do my columns and other writings.

The version I had was the Recorder Edition v9.5, and it required you to record your dictation either onto a handheld recorder, or directly onto a computer, before it would transcribe it for you.

I had been hearing good things on the Internet about the improvements in Version 10 of the software and I was very keen to give it a try because it would allow me to dictate straight into a wordprocessing program, doing corrections as I went along, and cut out the need to

record it first.

In December I started getting promotional e-mails from Nuance, the producer of Dragon, offering the Standard Edition v10 for only €49 (about R500). It sounded like a bargain to me and finally, after Christmas, I cracked and went online to order.

After a 12-hour download, and installing the program, there was a large disappointment in store. It turns out that the standard version of the software cannot import user files from previous versions and, even more astonishing, it cannot import its own user files in the

event that you have a hard drive crash or buy.

It should be explained that, to get the best out of Dragon, you have to train it to understand your particular voice and as you do so, and make corrections, your voice files are updated and the program gets more and more accurate. There is eventually a significant amount of time invested in your user files and the last thing you would want is to lose them and have to start the training all over again.

I immediately got on the line to the customer service office in Edinburgh and expressed my outrage. To my mind, it was unacceptable that I would not be able to save and restore my files in the future which, when you think about it, is a bit like Microsoft telling you you have to retype all your documents when you upgrade to a new version of Word.

The company eventually agreed to refund my money on the Standard Edition, let me buy the Preferred Edition at the upgrade price, and give me a small additional discount for my trouble.

Well, I wasn't happy at paying what would amount to double the amount I had made up my mind to spend, but I felt that I really did need to be able to save my user files periodically and import later, if need be.

So I went through the 12-hour download process again, and installed the new software. I then went through the training process by reading a section of one of the Dilbert books to it, and then turned it loose on the documents on my hard drive to allow it to analyse my writing style.

I was soon up and running and dictating in many different circumstances ranging from a text box on a web page to a window in Microsoft Word. My initial reaction was that the product was every bit as good as it had been built up to be and I found it much easier to

dictate and correct the text on the fly.

One of the main advantages of the Standard and Preferred Editions over the Recorder Edition that I had, was that it can be used to control the computer as well as for dictation. I am just getting used to the control feature and am still experimenting but, from what I can see, it would be more than a possible for someone to go through their day-to-day computer tasks without needing to touch a keyboard or mouse.

In addition to lessening the strain on people who use their computers a lot, I feel the software would be ideal to enable disabled people who are unable to use the keyboard and mouse properly, to use a PC every bit as well as a person not handicapped. This lesson was driven forcefully home to me in a video I came across while doing some research on the Internet.

Disabled blogger Jon Morrow explains the state of the nation as far as speech recognition software is concerned, gives a very good description of the sort of hardware you will need, and a brief demonstration of how he would use the software to create a blog post,

without ever touching the keyboard.

It's his opinion that Dragon NaturallySpeaking is the premier speech recognition software available today but he does say that both Windows Vista and Windows 7 have built-in speech recognition which, he

believes, is not too bad at all. By the sounds of it, it would at least allow interested people to give the technology a whirl.

I have seen Dragon NaturallySpeaking Preferred v10 on sale locally for R4 000 but, to my way of thinking, it would be criminal to pay that price when it is available for download at something just over a third

of that. This is a question that I've asked before but just what is it with South African retailers that make them so greedy?

- Links to the Jon Morrow video and previous columns are available at allan-fishnet.blogspot.com.

Related Topics: