Some people eat, sleep and chew gum, I do genealogy and write...

Thursday, February 17, 2011

A return to speech recognition

At the recent RootsTech conference in Salt Lake City, Utah, I attended a class taught by Anne Roach of FamilySearch. Right off the bat, she inspired me to go back and take another look at speech recognition. Off and on over the years I have spent time trying to use speech recognition software to speed up my writing. Unfortunately, each time that I tried to use the software it turns out that the software was not capable of following my dictation without an unacceptable level of errors.

Because of the volume of writing I have been doing recently, I decided to give speech recognition another try. I researched the programs, looked at the reviews and decided to purchase one of the newer programs. I had previously used  Dragon NaturallySpeaking and found it to be a very good program except for the limitations of speech recognition in general. This time, I decided to try the Macintosh version. The program is called Dragon Dictate and is based on the Dragon NaturallySpeaking model.

As I had expected, after receiving the program, I remembered that learning speech recognition software was very similar to learning an entirely new language. In this case, there are 184 pages of commands and instructions. Most of the instructions are fairly simple to remember but it is a challenge to learn to distinguish between the commands and the text you are dictating.

All of the speech recognition programs require a training mode, where the user reads a selection of text and trains the program to recognize his or her voice. It is critical to the use of the program that this training be done to the extent that the program needs to learn the user's voice or the speech recognition will be faulty. After spending a short time training the program, I decided to launch right into trying to dictate a blog post.

To my pleasant surprise, the program seems to work very well and is recognizing my speech without too many problems. The main problems seem to center around mis-recognition of the words and my inability to remember all of the commands. I would expect that as I studied the program and learn the commands that my ability to dictate will speed up considerably.  What I also found previously, was that I had a tendency to use the mouse even though I was trying to dictate with the microphone. Learning the commands will probably eliminate nearly all of the need to use the mouse.

I also found that having a good quality USB microphone was absolutely essential to adequate dictation. Even though the program came with a microphone, I purchased a higher quality microphone almost immediately and noticed a marked increase in the capabilities of the program to accurately represent the text that I was dictating.

For individuals with handicaps or other challenges that prevent using a keyboard effectively, voice recognition would be an effective way to enter text into the computer. Any time you are using the dictation program, it is very important to watch the text carefully as it is entered by the program so that you can make corrections “on the fly.” It is apparent from the dictation which I have all already made, I will have to go back and do a significant amount of proofreading, which is not one of my strong points.

It is certainly apparent that speech recognition has come a long way since the first demonstration that I saw at the world's fair in Seattle in 1962. However, I don't think that the speech recognition program will help me think or do research any faster, as it appears that this blog post took just about as long as it would have had I typed out the text rather than dictated it through the program. My goal is to continue to use the speech recognition software and from time to time, I will give updates on my impressions of how my use of the program coming along.


  1. Does it still require discreet speech - a brief pause between words - or have they advanced to the point where one can speak in normal fashion?

    I would be willing to bet that as you get used to it, and it to you, that your process time will decrease to where it may be faster than typing out the posts, assuming you continue to work with it a significant amount.

  2. On my Droid it has speech recognition to send a text message. What comes up sometimes is very funny.