[Noisebridge-discuss] Fwd: drama-o-meter

[Mike Schachter]

Hey MLers,

If anyone is interested in implementing a "drama detection"
algorithm, we should have an ML meetup about it. There's
tons of training examples on the archives of the noisebridge-discuss
list.

At first glance we could probably use tf-idf along with the network
structure of emails to do unsupervised learning, then train a supervised
learner on the manually-improved data.

http://en.wikipedia.org/wiki/Tf*idf

  mike



---------- Forwarded message ----------
From: Jake <jake at spaz.org>
Date: Tue, May 8, 2012 at 2:48 PM
Subject: [Noisebridge-discuss] drama-o-meter
To: "noisebridge-discuss at lists.noisebridge.net"
<noisebridge-discuss at lists.noisebridge.net>


can someone code up a thing that looks at the discuss mailing list traffic
and measures what portion of it is drama?  it should be easy.

a human will have to declare threads to be drama or not, and if they are,
and the total traffic for the week exceeds a certain percentage (say 50%)
the moderation automatically chokes off any further posts on any of those
threads until the percentage cuts back down.

i view the discuss list on the web portal like this:

https://www.noisebridge.net/pipermail/noisebridge-discuss/2012-May/date.html#end

and as you can see, it's really hard to find the hacking-related posts.

-jake
_______________________________________________
Noisebridge-discuss mailing list
Noisebridge-discuss at lists.noisebridge.net
https://www.noisebridge.net/mailman/listinfo/noisebridge-discuss