The goal of the Evidence Extraction and Link Discovery (EELD)
program is development of technologies and tools for automated
discovery, extraction and linking of sparse evidence contained
in large amounts of classified and unclassified data sources.
EELD is developing detection capabilities to extract relevant
data and relationships about people, organizations, and activities
from message traffic and open source data. It will link
items relating potential terrorist groups or scenarios, and
learn patterns of different groups or scenarios to identify
new organizations or emerging threats.
EELD is a concerted
effort to take the dramatic narratives of asymmetric warfare
(terrorism) as they are currently understood and make them predictable.
In the same way that an individual who has seen enough genre
cinema can quickly assess the arc of a given film's narrative,
including the motivations and attitudes of its characters, EELD
will take the "distributed script" represented by
phone calls, emails, web pages and other correspondence and,
literally, predict the plot.
Various agencies had
all the necessary elements to predict the events of September
11th, but they lacked the computational power and algorithms
necessary to resolve the narrative until it was too late. We
will not make this mistake again.
EELDs initial activities demonstrated the feasibility
of extracting relationships from text, and validated the detectability
of patterns representing terrorist groups and scenarios.
EELD has also developed two promising techniques for learning
patterns of activity, developed functional system concepts to
guide technology developments, selected techniques to develop
for evidence extraction, link discovery and pattern learning,
identified scenarios to validate the detectability of patterns
in unclassified and classified data, and initiated the collection
and characterization of documents for technology evaluations.
EELD's success is
entirely dependent on the continued development of Work-Life
scenarios that restrict action and dialogue to prescribed patterns.
Fortunately, general trends in business are moving towards more
restricted modes of communication, not only in terms of surveillance
but also in terms of exactly what gets said. As the language
of the workplace becomes more restricted and narrowed down to
exclusively commercial dialects, future implementations of the
EELD system will become more effective. Deviations from accepted
patterns of communication will become more readily discernable.
FY 02: EELD will develop and
demonstrate technology to extract relationships, and detect
and learn single-link type patterns.
FY 03: EELD will: 1) extend its capabilities to
the extraction of data from multiple sources (e.g., text messages
and web pages), with an ability to adapt rapidly to new threat
domains; 2) develop the ability to detect instances of patterns
comprising multiple link types (e.g., financial transactions,
communications, travel, etc.); and, 3) will develop the ability
to learn patterns comprised of multiple types of entities (e.g.,
persons, organizations, etc.) and multiple link types. 4)
begin a rigorous construction of the Islamic Strategic Narrative
to fit all parsed evidence into. It is recognized that future
safety cannot rely exclusively on evidence gathered in real
time. A forecasting system must be put into place that not only
takes economic and political factors into account (see Project
FutureMAP) but the overarching "storymode" of terrorism.
We must overcome the simplistic Crusades metaphor and dedicate
more efforts into truly undersanding the Grand Narrative that
the Muslim world is creating in the face of democracy and freedom.