Here's two cents from someone who has worked with transcriptions of conferences and interviews for
two decades now ...
There is no automated way to accurately transcribe the human conversation yet. Even a human
transcriber is usually highly inaccurate (at least at first) unless they were present at the time of
recording. We have used a professional court reporter for years to do our conference and interview
transcriptions and she is just fine with the 85% or so that is common language and words, but she
often mangled the 15% that is colloquial (sp?) or industry-specific or heavily-accented -- but since
she is a human she has improved impressively over several years. I've experimented with IBM and
Dragon software and even got a friend who used to work at IBM labs to run some material through
their most advanced software and the accuracy is worse than our faithful court reporter. By the time
you sit and edit all the mistakes and fix the punctuation and grammer, it's taken almost as long as
if you just had transcription skills and typed out the whole thing yourself.
I would suggest, for a large institution, or even a medium-sized one that can get some funding,
rather than pour tons of dollars into computers and software, best to find a skilled transcriber
like a court reporter looking for part-time extra work. Bring this person to your facility and teach
them about what you are doing. Listen to a few tapes with them and let them ask questions about the
content, accents, topics, etc. Give them a tryout and see how accurate they are. There will be a
point where you find the perfect person. I should also say that we've had good luck working with our
court reporter teaching her better punctuation and grammar practices, which she said has made her a
hit with the lawyer community and gotten her better gigs from them too. Also, over time, she is more
familiar with our industry and is making fewer and fewer howlers with industry-specific language.
The last set of trascriptions she did of long on-stage interviews from one of our conferences, the
editing time was so much shorter than the first few that we noticed and paid her a bonus for saving
One man's experiences ...
-- Tom Fine
PS -- no reason a professional transcriber couldn't also learn how to enter metadata information to
go with the transcription.
PPS -- in case it's not painfully obvious, I'm saying human ALWAYS beats machine if any sort of
skill or nuance is required! Transcribing human conversation is a great example of skill and nuance,
more so if the recording quality is not top-notch.
----- Original Message -----
From: "Andy Kolovos" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Wednesday, July 30, 2008 7:49 AM
Subject: Re: [ARSCLIST] Transcription/Close Captioning Software for Oral History Recordings
> Depends on what you mean by "time-stamped". If you want to basically do your own "Docsoft-style"
> thing, you got me. If you want to transcribe with a system that will allow you to manually add
> accurate time codes into the text, we've been using Start-Stop. Express Scribe might do this as
> well, but our transcriptionist greatly prefers Start Stop to Express Scribe. Still, I get the
> impression that you're trying to do something else here.
> The Docsoft system seems interesting, but looks expensive and proprietary--all the same, I hadn't
> encountered it before so I appreciate the reference.
> Have you posted this to the OHA oral history list?
> Markus Wust wrote:
>> <div class="moz-text-flowed" style="font-family: -moz-fixed">Dear list members,
>> I am looking into using transcription or close captioning software (e.g., Docsoft) to create
>> time-stamped transcripts of oral history recordings. Are you, or do you know someone who is using
>> such an application for this purpose? If so, what has been your experience?
>> Thank you,
>> Markus Wust
>> Digital Collections and Preservation Librarian
>> North Carolina State University
> Andy Kolovos
> Vermont Folklife Center
> 88 Main Street
> Middlebury, VT
> (802) 388-4964 [voice]
> (802) 388-1844 [fax]