I received a question this morning from newsman Rick Smith that I cannot answer. Certainly, some among our readership of several hundred writers can shed some light on it:
Tom,
How does Dragon Naturally Speaking work when you tape record an assortment of different people? Can the software turn that into copy pretty well, or is it really set up just to work with your own voice. I’m a news reporter; I tape a lot of interviews; and I’d love to be able to download them and have them turned into copy. — Rick Smith
Rick has a good question. I know that our loyal reader Michael Downend uses Dragon Naturally Speaking, but he has spoken about the necessity of “training” the software.
The topic has come up before when I wrote about the Olympus Digital Voice Recorder, a post that prompted a number of comments. Michael, can you provide more insight into trasncription software? Anyone else? Just click on the comments link below to join the discussion.
Sorry, comments are closed.
Comment by
Tracy Urban
1 Jun 2008
Sadly DNS only works with your voice. I tried to run an interview through it and only got garble.
It’s pretty good software to use for dictation (I sometimes dictate my email replies if I’m tired of typing) but I can’t recommend it for interviews.
Comment by
brettbum
2 Jun 2008
I use DNS all the time and test it out in a number of odd scenarios. Trying to capture a group of people would be out of scope for the device, however if you had good microphones to maybe 2 people and recorded the sessions at 48kbs or better, you might be able to teach the software to understand the words of the same 2 people over and over again.
Similarly, if you conducted an interview with yourself (a person that DNS is already familiar with) and another person, and that interview is long. Say more than 10 – 15 minutes. You might be able to transcribe the first 10 minutes to teach the software the second person’s voice and then rerun the recording through transcription at a higher rate of accuracy.
These are both things that I intend to test soon, right now I’m just offering up my hypothesis. I wouldn’t spend any money on the software to achieve goals based on these hypothesis, but I do find the software very very useful myself.
Comment by
Tom Colvin
2 Jun 2008
Tracy and Brett: thanks for sharing with us the benefit of your experience. I’m not surprised at your reports, but I have had no experience in the area and appreciate your willing to take time to share. That’s what this blog is all about.
Comment by
Michael Downend
4 Jun 2008
DNS is not for multiple speakers. I’ve used it for years and found it invaluable for its intended purpose: single-speaker use. As your previous correspondents have said, you’ll waste your time attempting to use it for voices for which it has not been “trained.”
MD
Comment by
Leisureguy
6 Jun 2008
My wife works at Stanford and just this week explored this question. She talked to a guy who had tried it. It doesn’t work: the variety of voices confuses the software. Best use the software just for one person at a time.
Comment by
Tom Colvin
6 Jun 2008
Thanks to all of you for your input. I think that together we’ve definitely answered the question. Just hope that Rick Smith has had a chance to learn from your advice.