> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On
> Behalf Of Chad Cooper'
[snip]
> Nick, different subject: have you considered trying to gain metrics on
> discussion threads based upon theoretical social boundaries, to
> come up with
> a "Quotent" or matrix of ...tolerance, intolerance, diversity, or
> commonality of values?
I'm not sure how I'd measure such things. However, it's not difficult to
spot flame wars, so I can generate a "heat index" for communities. I'm also
measuring connections among communities, which might be an indicator of
openness to different ideas, if one assumes that a community with strong
connections to "distant" topics (in terms of subject focus) is populated by
open-minded people.
I'm not sure if I've said so here, but lately, I'm focusing on finding
diverse points of view on a given subject, using a combination of traffic
and text analysis. The traffic analysis spots people with influence; the
text analysis looks for dissimilar language among the influentials. A lot
of other things have kept me busy, but I'm aiming to get something
accessible via the web in the next week or so.
> Or is this one of those contexual problems in processing mail threads to
> understand the context of the message? I am just thinking out loud here...
> bear with me...
I think it's possible to categorize messages by point of view with 80
percent accuracy or so, although that surely would vary by how fine-grained
one wants to be. That's not to say that these points of view can be
labeled, just that they can be differentiated by the words, phrases and
references offered. Text analysis is very tricky, though. Negations are
very slippery to identify and words often are used in meanings that are
exactly the opposite.
Nick
