About two months ago, I ran across a blog post by Jon Udell of InfoWorld about his experience with Dragon NaturallySpeaking 8. Like him, I toy with speech recognition technology every few years to see how it has progressed. After reading his post and watching his video, I decided to order a copy of the Standard edition for myself.
The software comes bundled with a headset that includes an attached microphone. I’m not sure of the headset quality, but so far, I’ve been very happy with its results. Noise from the surrounding environment always wreaks havoc with speech recognition applications, and a good microphone can make a significant difference (especially if it’s a noise canceling microphone).
The software installation was very smooth, and before long, I was reading through the included training texts. For some odd reason, I enjoy reading the training text for these applications. Perhaps it’s because I know that the more I train the software, the more accurate the recognition will be. Some of the texts are relatively short, but a couple of them took me more than 45 minutes to read through.
To further increase accuracy, NaturallySpeaking 8 can optionally scan through your documents and e-mail to learn your writing style. Although I don’t know all of the technical details, I’m guessing that it looks for unique words, proper nouns, and spellings so that it can recognize or suggest them later. Pretty cool.
After setting everything up, you can fire up your favorite application, turn on the microphone, and begin dictating. As you speak, NaturallySpeaking 8 listens to sentence chunks and uses context and grammar rules to figure out what you said. Best of all, there’s no need to talk in a stilted manner, and you don’t have to insert pauses between your words. You can actually speak naturally. Imagine that!
So you could see how I created this blog posting, I downloaded an evaluation copy of Camtasia Studio and recorded myself dictating this text into Microsoft Word. In the interest of full disclosure, I did write this post ahead of time, because I didn’t want to fumble around for the demonstration.
All in all, I’m very impressed with the accuracy of the recognized text. The problem is that I type very quickly, so I’m not sure it saves me any time. However, for getting thoughts into the system or for people who don’t type for a living, this is a great application.
Update: As you'll notice in the recorded video, there were a couple errors that I had to fix manually for this post. Also, you may need to install the TechSmith Codec to propertly view the video.
It is also really good for putting numbers in spread sheets and such. Unfortunately, I had a computer crash and I couldn't get my dragon naturally speaking to reinstall. ( The file SAPIDLL.DLL did not register itself. - was the message I kept getting. ) Checked at their support web page and they had a patch program (spcapi) that solved that problem but now their website will not generate a serial number for the reinstall saying invalid key. (Which it isn't as I downloaded what they had registered for me.)
Here is the interesting part. They want to charge me either $19.95 for phone call or $9.95 just to send them an email!
So my report would be good product - TERRIBLE SUPPORT. All to common a problem in software circles.