Voice Recognition

Looking for a great tool for capturing info on the go? Try a voice recorder with voice recognition software.

A while back I was going to the largest trade show in our industry, and the first show in my new CI position. It’s very long, very big, and I knew I would be learning lots of stuff. I also knew a voice recorder is a great tool for storing the tidbits and observations as they happen – it’s great because it’s easy to record all sorts of stuff that somehow never gets written down when taking notes. But I wasn’t looking forward to transcribing everything each night.

So I decided to try voice recognition again. I had tried it several years before, when I was an engineer. The real problem back then was recoding quality and the noise level in the factories where I was using it. In short, it didn’t work very well. I figured things would be much better after 7 or 8 years of development, and they certainly are.


During my last foray I learned that the quality of the recording was critical, so I started looking for the best voice recorder I could find. I finally decided on the Sony ICD-BM1, bundled with Dragon Naturally Speaking software. Some folks are very fond of the Olympus gear. The key is to get the highest quality you can afford. Microcassette recorders just don’t work for this. You can use the newer Palm PDAs with Audacity software, but the Dragon site implies lower accuracy with this outfit. The total cost for the equipment was about $325.


Before you can use the system, you have to train it for your voice. The training needs to be done in a quiet place. Speak the way you will in actual use. I first did it during a long drive with my wife, and it didn’t work because of background noise. The system tolerates much more noise in actual use, however. Also make sure you hold the recorder close, and in a position you can easily duplicate – it affects the sound level, which in turn affects transcription accuracy. While it can be tedious, I learned the hard way that training is critical to making an effective system.

Actual use

The first task was dictating questions for scavenger cards. I was able to dictate 450 cards in a total of 90 minutes. This was a lot faster than typing, and also easier because the questions came from looking through a lot paper work. Syncing the recording via USB was snap with the included software, which also has some nifty features for manual transcription. Then I started the transcriber software, which dumped the transcribed text into a special text editor. There were lots of errors, but I used the “correct this” feature to correct them, which also teaches the software how to interpret my speech better. I transcribed again and voila! The transcription was already much better. A few more corrections, and it was transcribing pretty darn well. The trick is to record a little, transcribe and correct, record some more and repeat. Start with small recordings, and increase the size as the accuracy improves. The system learns fairly quickly, and I also learned to improve my speech.

Ok, so it worked very well while sitting at my desk. How would it do while walking a trade show floor? More than well enough for my purposes. I transcribed and corrected each night in my hotel room, and by the third day the only mistakes I was getting were totally new names, or instances when I misspoke. In fact, that was the biggest problem – I didn’t realize how fractured and goofy I spoke. Thinking about what I want to say and organizing it into sentences before I speak helps a lot. The more I use it, the better it gets.

At the show I gathered over 20 single-spaced pages of notes, and while it did have to be proofread and some corrections made, it was lot less work than manual transcription. The noise level wasn’t a problem unless it was very high, like standing next to a noisy machine. I got in the habit of keeping the recorder with me, and it was invaluable for storing all the things I heard during dinners, travel and other times as well.

If you are looking for a very handy way to capture all of your notes and observations when paper and pen just don’t work, give voice recognition a try. While it was expensive and somewhat tedious to get started, it has become an invaluable tool and paid for itself more than once.

3 thoughts on “Voice Recognition

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s