Database Reference
In-Depth Information
of the Dabhol Power Company (DPC) in the Indian State of Maharashtra.
Whereas the company's efforts in India had been ongoing for several years,
emails of the first six months of 2001 reflected several of the day-to-day deal-
ings with that situation.
By October of 2001, Enron was in serious financial trouble. A merger
with the Dynegy energy company fell through and forced Enron to file for
Chapter 11 bankruptcy. Many of the emails in the months of October and
November were newsfeeds from various organizations that were being routed
through the company. As it was reported that Chief Financial Ocer Andy
Fastow was heavily involved with the deceptive accounting practices,
4
it is not
surprising that a topic we labelled
Fastow companies
emerged. Predictably, a
college Football
topic emerged in late fall as well. One of the surprise topics
uncovered was an education-related topic due in large part to the interests
and responsibilities of Vince Kaminski, head of research. Kaminski taught a
class at Rice University in Houston in the Spring of 2001, and was the focal
point of emails about internships, class assignments, and resume evaluation
(5).
Since only eight of the 25 topics had any discernible meaning, it would seem
apparent that a significant amount of
noise
or undefined content can still per-
meate a term-author-month array. In some instances, there are indicators of
a possible thread of some kind (not necessarily directly related to Enron), but
a closer inspection of those emails reveals no identifiable topic of discussion.
The daily results reported in (5) provided a similar interpretation as the
monthly results but at a finer resolution. In general, one observed four dif-
ferent types of discussions: (
i
) discussions centered largely on one or a few
days, (
ii
) continual activity, represented as multiple weekly spikes in activity
throughout the year, (
iii
) continual activity with lulls, where a period of calm
separates bursts of discussion, and (
iv
) a series of weekly spikes of activity
usually spanning three or more months.
Of the 25 discussion groups mined with the PARAFAC model, roughly half
were of the first type. Examples include a flood of emails about the possible
Dynegy/Enron merger (November 11 and 12th), a topic on January 7th in
which Enron employees (Kean, Hughes, and Ambler) were discussing India
based on an article published by Reuters and another media report, and a
discussion centered on the August 27 U.S. Court of Appeals ruling on section
126 of an Environment Protection Agency code.
The nonnegative PARAFAC model identified temporal patterns similar to
those of PARAFAC with a majority being a series of weekly activity spikes
spanning three or more months. Roughly one third were single spikes pat-
terns, and just two discussions are somewhat bimodal with a lull. A few of
the more interesting (single spike) discussion groups extracted by the nonneg-
ative model included a flurry of emails on August 22 in response to an email
4
Setting up bogus companies to improve Enron's bottom line, for example.
Search WWH ::
Custom Search