Defining groups of people based on
keywords
found in emails and other communications
A Social Visualization project by Michael Hines
meOwns uses term frequency-cosine similarity to determine how similar two friends are based on descriptions of the items the two users own.
About their algorithm | The meOwns web site
The original paper describing what is now known as tf-idf
Judith Donath and Danah Boyd both present similar ways of determining social networks by email habits, however their methods look primarily at the headers of emails rather than the body.
Boyd's master's thesis is here | Donath's draft of Visualizing E-mail Archives is here
The mail_receiver daemon is heavily based on the following link: Gmail mail_receiver daemon
The Flare docs and demos were infinitely helpful, as this was my first Flex project, let alone Flare. UC Berkeley's Flare