Abstract

I present a dynamic method for visualizing a conversation-based network built from IM logs and emails based on ongoing conversations. Conversations are analyzed for keywords and then graphs are built connecting people to keywords creating a dynamic representation of the underlying social network.

Motivation

Conversations are not always unique. Often times they are threaded and comprised of many sources. People are rarely linked only by other people. Often times friends meet through people, but the friendship is more than just a bridge of people. Commonly friends and acquaintances form because they tend to enjoy discussion of the same interests, or sometimes they communicate because of necessity of work. I show linking people by keywords is better than through the mere knowing of another person.

Project

The project described here is a graph-based visualization. Using a force-directed graph system, with edges as springs and nodes as positively charged particles, we are able to automatically determine links between people. These links are useful information in themselves as the links are defined by shared keywords.

Read more in the original project proposal