Quickly Locating Groups and Group Leaders with E-mail Analysis

Researchers at Hewlett-Packard have found a way to quickly identify "communities of practice" within an organization -- that is, people who have similar beliefs and goals and work together to achieve them. Further, the researchers' algorithm, run "in a matter of hours" on a desktop computer, can find leaders of these groups. This excerpt from the paper's introduction, contained in this .PDF document (296KB), do a better job than the abstract (available in HTML here) of explaining understandably what they've achieved at HP:

. . . The method uses email data to construct a
network of correspondences, and then discovers the communities by partitioning
this network in a particular way . . . . The only pieces of
information used from each email are the names of the sender and receiver (i.e.,
the ?to:? and ?from:? fields), enabling the processing of a large number of emails
while minimizing privacy concerns.
We describe an experiment performed within our own organization, HP Labs,
using nearly one million email messages collected over a period of roughly two
months. The method was able to identify small communities within this 400-
person organization in a matter of hours, running on a standard Linux desktop PC.
In addition, we utilized the network of correspondence to identify leadership
within these communities.