Study of topics and thread participation

[sbenthall]

First, generate a topic model from several mailing lists.

Next, test with K-fold validation a classifier trained on the topics of the first message in each thread to guess whether the thread will be long or short in (a) # of messages, and (b) # of participants.

Test the accuracy of this classifier with a range of hyperparameters to the topic model.

Write this up in a Notebook with conclusions about whether topics can be used to determine thread activity.