At 10:27 AM +1000 8/18/00, Robyn Henderson wrote:
>Having just joined this list, I have no idea what discussion has been going
>on. However, I'm hoping that someone will have some information to help me.
>
>I'm about to start interviews as part of my data collection and would like
>to find the easiest way of transcribing them. I plan to use Fairclough's
>minimalist approach to transcription, which means that I'm not concerned
>with recording pause times etc. So, has anyone used any voice recognition
>software that copes well with interviews?
>
Yes, we have thrashed this topic just about to death several times. Here is
my take. I use Dragon Systems NaturallySpeaking and it works okay. It takes
a lot of training, a lot of patience, and a really fast computer with vast
memory (my 250 mhz AMD K6 with 96 mbytes was not up to the task). I had to
upgrade my hardware in order to get satisfactory performance and avoid
inducing apoplexy in the user (me).
The technique is to listen to the recording, preferably with a
transcription machine so you can precisely stop and start the tape. My
original plan was to listen and speak simultaneously and the software would
keep up. A better plan was to listen to a sentence or two, stop the tape,
speak what I just heard, correct it if necessary, then proceed.
Using this technique, I can transcribe a one hour tape in about two hours
or so. This is a little faster than a good transcriptionist can manage and
only cost me around a thousand US dollars for software and hardware upgrade.
This all sounds awful, but I am actually pretty satisfied with the results.
I figure I can amortize the cost of the hardware upgrade, so it is not fair
to charge it all to NaturallySpeaking.
Now, as to minimalist approach to transcribing. This depends on your
research problem and what you plan to extract from the data. I would
recommend transcribing everything you can stand, on the logic that it is
easier to leave something out later that you don't need, than it will be to
go back, find something you do need, and edit it in. Of course, if you are
going for some sort of semi-quantitative approach, such as content
analysis, then a minimalist approach might work just fine.
Now, if I have said anthing stupid, blame it on drinking Irish whiskey
while listening to the Austin Lounge Lizards at 11:30 on a Thursday night:
their latest album, not in the stores yet, _Never an Adult Moment_.
Elliot Richmond
PhD candidate in science education
University of Texas at Austin
[log in to unmask]
http://www.geocities.com/Athens/4758/
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|