An injury to my arms led me to start using IBM most recent
voice-recognition software as a means of inputting all my text. And I must
admit to being very impressed. I can certainly produce text much more
quickly than I can type, particularly with my damaged arms. The training
period was quite quick but it is now only after a month of using that the
real power of the programme is becoming apparent.
I have been using this to complete the help files for my new programme
CISAID (Code-A-Text Integrated System for the Analysis of Interviews and
Dialogues). One of the features of CISAID is that it allows text to be
input as if typing into a word processor. There are a variety of different
methods for coding and analysing the data. One feature is a transcription
system which works either via the mouse or the keyboard. It has all the
features of a normal transcription system including the ability to slow
down speech, insert periods of silence and control volume.
I have been using this system in conjunction with voice recognition
software and I am impressed by the potential. Having copied my media file
on to the computer I can click on one button to play the media File stop it
and then dictate what has been said directly into the CISAID Text window.
In fact it should be possible to transcribe a complete interview without
touching the keyboard.
One of the problems with voice-recognition software has always been that
many para verbal sounds cannot be reproduced. However the use of voice
macros, which allow you to associate a specific piece at text with a sound,
can allow you to insert para verbal utterances by dictation. (It
incidentally also allows the insertion of a annotation codes which are
embedded in the text by CISAID. These alongside the open-ended comments can
be inserted via the voice-recognition software. CISAID keeps two forms of
index one of the words in the text and another of the comments and
annotations; meaning that you can search along these three dimensions as
well as on Rating Scales.)
The system is not perfect but certainly I would expect it to eventually
reduce the transcription time for me by at least 50 per cent and for a
slower typist much more. However, there is always a downside and in this
case it is machine specification-I already have 256 megabytes of Ram on a
450 MH computer and this is not enough to really get most from the
procedure. It is not machine speed which is critical but the memory
required to run two powerful multimedia applications in parallel.
However, what this seems to indicate is that we are in sight of methods
which will make the transcription of text much easier. However, there is
one further difficulty -namely when I repeat what I hear on the recording
I tend to convey the emotion in my tone of voice and that confuses the
voice recognition software. The software is very powerful but only within
specific limits.
For those investigators who wish to include an understanding of features
such as tone of voice CISAID will be able to generate vocal profiles
consisting of volume, pitch and speech rate on a segment by segment basis
and automatically add these to the codes.
Regards
Alan Cartwright
and8/01/2000 +0000, Elliot Richmond wrote:
>At 3:27 PM 1/28/2000, Mark Priestley wrote:
>>b) using speech recognition software
>
>I can only respond to this item in your message. I am trying out the Dragon
>Systems NaturallySpeaking Mobile Edition. It comes with a digital recorder
>on which one can record memos and later download those memos to the speech
>recognition software.
>
>So far, I am not impressed. It may be that my computer (250 mHz AMD K-6
>with MMX techynology) is not up to the task. Right now, I am going through
>*all* of the training selections and also building the vocabulary to see if
>it helps.
>
>No speech recognition software will recognize multiple speakers in the same
>session. My NaturallySpeaking software allows multiple users for different
>sessions, but each session is single-user.
>
>I am still satisfied with the performance I get by listening to the tape
>and repeating the words back as they are said. I do this with video tapes,
>audio tapes, even printed documents (although the OCR software with my new
>scanner does a great job). I think a foot pedal controlled transcriber
>would be a great help, but so far I haven't been able to convince myself
>the cost is worth it..
>
>Elliot Richmond
>PhD candidate in science education
>University of Texas at Austin
>[log in to unmask]
>http://www.geocities.com/Athens/4758/
>
Alan Cartwright PhD
Code-A-Text Developer
Email [log in to unmask]
web page http://www.code-a-text.co.uk
Also
Senior Lecturer In Psychotherapy
Kent Institute of Medicine and Health Sciences.
University of Kent. UK.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|