[Air-l] Abstract: Measure of Discussion Activity in Threaded Discussion Spaces

Avner Caspi avnerca at openu.ac.il
Tue Aug 13 00:22:34 PDT 2002




-----Original Message-----
From: David Wiley [mailto:david at wiley.ed.usu.edu]
Sent: Sunday, August 11, 2002 4:07 PM
To: 'air-l at aoir.org'
Cc: david.wiley at usu.edu
Subject: RE: [Air-l] Abstract: Measure of Discussion Activity in Threaded
Discussion Spaces

Thanks for your feedback! I'll try to address these issues all at once.
First, as for the semantics of the d value, d is the average reply depth;
in other words, d simply tells you what level of reply the average message
is at. So, for the tiny thread:

A (Reply depth of 0)
-B (Reply depth of 1)
--C (Reply depth of 2)

The unadjusted reply depth would be (0+1+2)/3 = 1. The paper presents some
adjustments to fine tune this value, which almost always lower it. So, the
difference between d=13 and d=2 is that the d=13 thread would have to look
like this:

A
-B
--C
---D
<snip>
-----------L
------------M
-------------N
--------------O
--------------O
--------------O
with many more 14th level replies for the value to ever get over 13, while
a d=2 thread looks much more normal, something like:

A
-B
--C
---D
---D
---D
---D

d=13 tells you that one group is discussing something to a much more
significant depth than the group whose d=2.

As for the value seeming to grow in size as the number of messages grew,
this should not happen as d is intended to be an average, and should
therefore be unaffected by the size of the archive. The discussion
in the paper is obviously not be clear enough (which is why I asked AoIR
for feedback!). I will try to fix this. Also, would you be willing to send
me the archive whose d=13 so that I could run it through our software
here? I would love to see it!

Perhaps it is domain specific, but in newsgroups dedicate to humanities
you'll find very deep threads. Two examples I digged: alt.culture.fyrom and
alt.news.macedonia . I checked your thumbnail in Hebrew public newsgroups
(philosophy, to that matter), therefore I can't send you that archive.

 
Have only just skimmed this but it looks excellent! Our goal was to create
a value that provides a thumbnail of the level of discussion activity
whose calculation is easily automated, so that it could be integrated into
Course Management Systems and the like. It looks like this paper you've
referenced could help improve our calculation. Many thanks for this
pointer and your other feedback!


Avner 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.aoir.org/pipermail/air-l-aoir.org/attachments/20020813/340dc4ce/attachment.html>


More information about the Air-L mailing list