The authors work on the problem of identifying subgraph bursts in this structure, consisting of a significant number of link appearances among a small set of sites in a short period of time. They first extract dense subgraphs from the blog graph, which correspond to all potential communities (whether or not bursty). Given a specific community the relevant events correspond to the arrivals of the edges in the community. Then they extended the hidden state model to discover dense periods of "bursty" intra-community link creation for each extracted community in the graph.
The key references are [13] which detected communities by enumerating all bipartite cliques (up to a certain size) in the web graph, [11] in which an infinite automaton was proposed to identify bursty features and their bursty structures.
Reviewed by
EvanXW
- 2009-04-18 08:19:41
Search
Reviews
[Write a review of this article]
Find related articles from these CiteULike users
