Susanne has got me thinking about the implications of overlapping codes and
the different conceptual frameworks implied and I think an example may help
others see the differences in approach more clearly. Below is the
pertinent section from her message.
>I suspect the problems of overlapping codes comes from the
> > different roots
> > from which C-I-SAID evolves. My understanding is that many qualsoft
> > programmes allow you to apply a code or a label to a segment and
> > then bring these together into families etc.
>
>that is not quite correct - what is meant by overlapping codes is that
>you can code let's say a segment from time 1 min 30 sec. to 1 min 50
>sec. and an overlapping segment starting at time 1 min 45 sec. - 2 min
>30 sec. This is what I meant with the exlusive or either/or coding in
>CISAID.
In fact in C-I-SAID the problem is one of segmentation and when this is
dealt with the issue of overlapping coding is less problematic. I shall use
an example which is a schematic abstraction from a real project of which
there is more detail in the C-I-SAID user guide.
The basic data was a psychotherapy interview. The interview was divided
into speech units (essentially speaking turns) and I became particularly
interested in two types of therapist behaviour - friendly and hostile.
Within the interview you might find a speech unit with three sentences.
Sentence one 'friendly'
Sentence 2 'hostile and friendly'
Sentence 3 'hostile'.
In an overlapping coding model you could create a friendly code (attached
to sentences 1 & 2 ) and a hostile code (attached to sentences 2 & 3).
In C-I-SAID you would divide the speech unit into three segments the first
coded on a friendly scale, the second on the friendly and hostile scale and
the third on a hostile scale. Thus by re-segmenting the overlap disappears.
This coding can now be processed as a single speech unit with multiple
codes or as three separate segments. By using this method in C-I-SAID it
then becomes possible to do many other things such as ask questions about
the similarities and differences between these three types of segment (they
turn out to be distinct in terms of content and acoustic properties) and
their impact on the client.
This actually indicates a wider difference in approach because in C-I-SAID
codes are primarily attached to segments and segments are defined before
codes are applied. So to undertake this very detailed analysis you would
normally define your segments first. In this particular case the segment
would be defined as a distinct thought unit (normally a sentence) and the
codes would be attached to the segment. There are many built in methods
for automatically defining segments such as by words or grammar or time for
media files. Or they can be defined by the user.
In the overlapping coding method - it is the code which appears to be
defining the unit to which it is attached. You can do this in C-I-SAID
using annotations. As such they are useful for clarification and searching
but you cannot process them much further.
C-I-SAID is designed for very detailed analysis of documents whether they
contain text and / or media. The documents can be quite large and 1000
segments from an hours speech is not uncommon. Quantitative processing
thus becomes important. Really all this does is to classify, describe and
contrast the segments and whilst it does present statistics it also
presents tables and charts. Click on a data point in a chart and you are
back to the source data.
The irony is that while this may seem a very positivistic approach it
actually can function in quite different ways and I often feel, that by
using content analysis particularly, one can achieve a more grounded
description of a document or group of documents - and more quickly than is
possible with manual methods. The quantitative analysis highlights the
areas of interest and because it is closely linked to the source data it
enables the investigator to become aware of patterns that might otherwise
be missed.
Alan Cartwright PhD
Developer Code-A-Text MultiMedia Products
Hon. Senior Lecturer Kent Institute of Medicine and Health Studies.
Email [log in to unmask]
CISAID: Powerful Multi-Media Software for Analysing Interviews and Dialogues.
CTANKS: Word processing, Recording, Transcription, Searching and Report
Generation in a single user friendly package.
Information at
Code-A-Text Web Page <http://www.code-a-text.co.uk>
|