Once I had the first Blogging Heroes interview in the can, it was time to face the inevitable: transcription. Transcribing someone else’s words is tedious for writers. We’re accustomed to pulling words out of our heads without routing them through our ears first. That extra step is time-consuming and often confusing. And the physical process of transcription—listen, pause, back up, listen, type, listen, back up—can be slow and maddening.
But several people told me that audio interviews could be quickly and easily turned into text files with a voice-recognition program like Dragon NaturallySpeaking. Hmm…there was that “easy” thing again. I was suspicious. I figured only government outfits like the NSA had voice-recognition software that good. Still dubious, though hoping for a miracle, I sprang for a copy of NaturallySpeaking.
NaturallySpeaking is indeed an excellent program. It does everything Dragon Software promises, and does it well. But they never promised that it would recognize more than one voice at a time. It transcribes any voice that it’s been trained to recognize splendidly. But it handles only one voice at a time.
Faced with the tedium of typing and having spent $200 for the best voice-recognition software available, I still hoped for a shortcut—a way to get the words from audio to text format without pounding them into the keyboard. I asked around a bit and found it. A blogger named Dan Broadnitz suggested dictating the interviews into NaturallySpeaking as I listened to the recordings. I tried it. I donned a headset-with-microphone and played an interview with Sony’s Digital Voice Editor software (included with my Sony recorder). As I listened to the interview, I echoed back the subject’s responses into the microphone.
It worked! Trained to my voice, NaturallySpeaking faithfully transcribed the interviews. No stopping, no backing up. If the speech was too fast for me to echo, I slowed the playback. Quite often, just a few words would jog my memory of an interview enough that I could repeat entire sentences before I heard them.
Once the interview text was in place, I cleaned up the transcription errors (10 to 15 percent of the text). Then it was time to polish the text, culling out hesitations and misstatements, getting sentences into shape, and combining related sections of the text. I was careful to preserve the meaning, vocabulary, and speech pattern of each individual.
(Note: Only one of the interviews in this book was conducted by e-mail, at the interviewee’s request. See if you can figure out which one it is.)
P.O. Box 175, Oxford, OH 45056
Copyright © 2007, Michael A. Banks