Comparison Mrinmaya et. al. WWW2012 and McCallum et al 2004

Papers

The papers are

• Mrinmaya et. al. WWW'12
• The Author-Recipient-Topic Model for Topic and Role Discovery in Social Networks: Experiments with Enron and Academic Email

Comparison

Both papers deal with the problem of community detection in social networks (A major chunk of McCallum's paper mainly deals with conditioning topics on both the sender and recipient, which do not have much to do with community detection. But the second half of the paper about the RART model deals with modeling communities, which they refer to as roles).

In essence, the two papers are similar in the sense that McCallum's role in the RART model can be thought of as Mrinmaya's definition of community in the TURCM model. The authors in McCallum's paper test their model on the Enron email corpus and the McCallum email corpus, whereas the authors in Mrinmaya's paper test their model on the Enron email corpus and Twitter data (prepared by Mrinmaya). For emails, there is only one type of email (just plain email), but for Twitter there are different types of interactions that you can do (i.e. normal tweet, retweet, @reply, mention, etc...). Thus, Mrinmaya explicitly models the type of interaction into the model (which is represented by the latent variable $x$ on the left in the figure below). Other than that, the big idea behind the RART model and the TURCM are similar.

However, one slight difference between the TURCM model and the RART model is in how the community and the recipient is generated:

• In the TURCM model, for each user $u$'s post $P$, a community $c$ is chosen, and based on $c$, the recipient $r$ is chosen. (Thus, the community is assigned with respect to a user's particular post, not to a particular user. But it is possible to marginalize over all posts of a user to get the community distribution for a user).
• However, in the RART model, both the author and recipient's community assignment is based on the users themselves.

Questions

1. How much time did you spend reading the (new, non-wikified) paper you summarized? About 1 hour
2. How much time did you spend reading the old wikified paper? About 2 hours
3. How much time did you spend reading the summary of the old paper? About 5 min
4. How much time did you spend reading background material? About 30 min
5. Was there a study plan for the old paper? Yes
   1. if so, did you read any of the items suggested by the study plan? and how much time did you spend with reading them? The first 2 were papers I have read in the past. I did a quick read (mainly on introduction) on Airoldis' Mixed Membership Stochastic Blockmodel paper, which took me about 30 minutes.
6. Give us any additional feedback you might have about this assignment. '