We (actually, one of the guys working for me) have been
working with tools built on Latent Semantic Analysis (LSA)
to map threaded discussions and provide some convergence to
apparently disconnected threads. The result is better
understanding of the topic as measured by tests of the
students afterwards... The fundamental concept is that you
measure "similarity" of two or more messages in a lexical
hyperspace. And similarity is not a yes/no like pregnancy.
And no, I don't have references to the publications. I'll
try to dig it out at the office next week if anyone wants to
follow up.
Stan
Jostein wrote:
----- Original Message -----
From: "D. Glenn Arthur Jr." <[EMAIL PROTECTED]>
I guess once something is off topic it's off, but there may be
a concept of "further _from_" the topic once you're already off.
But I'm not sure what the metric would be.
Of course, once you have a topic-space and a metric for
distance-from-topic, together they define a topology,
n'est-ce pas?
Okay,
Sounds like you guys need a map to navigate by.
Try here:
http://www.ontopia.net/
:-)
