Motivation

1
Trust on Blogosphere using Link Polarity
Anubhav Kale, Akshay Java, Pranam Kolari, Dr
Anupam Joshi, Dr Tim Finin

• Motivation

Link Polarity Computation
Experiments
? Test Dataset from Buzzmetrics 3 contains 12 M
post-post links and reference dataset from Adamic
et al 4 contains 300 blogs labeled as left and
right leaning. ? Goal is to classify blogs in
Buzzmetrics
1 . Can you track the buzz for iPod in blogs ? 2.
Can you find the blogs that are iPod fans and
iPod haters ? 3 . In general, how can you target
the right set of individuals - like-minded
blogs for advertising ?
Trust Propagation
1. Guha et al 1 model based on applying atomic
propagations iteratively. 2. Mi1 Mi Ci
Perform till convergence M Belief Matrix Ci
Atomic Propagation Ci M MTM MT MMT
Problem Statement
Convert a sparsely connected non-polar blog
graph into a densely connected polar graph with
sentiments across each edge and use the polar
graph to model trust.
Approach
Direct
Transpose

• Sentiment detection to determine polarity
  of blog-blog links
• Trust Propagation to create polar links between
  blogs having no explicit links
• Label blogs as left or right leaning based on
  their polarity from influential blogs

Co citation
Coupling
1 Guha et al - http//citeseer.ist.psu.edu/guha0
4propagation.html 2 http//www.pacificviews.org/
weblog/archives/001989.html 3 Buzzmetrics -
http//www.nielsenbuzzmetrics.com/ 4 Adamic et
al - http//portal.acm.org/citation.cfm?id1134271
.1134277