Natural Language Sanitizing/Downgrading

Page Content

Victor Raskin - Purdue University

Sep 18, 2002

RealVideo

Abstract

Sanitization/downgrading/declassification is the task resulting from the
need to propagate private, classified, closed information down to public,
unclassified, open systems while avoiding any "data spills," the divulging
of sensitive information to unauthorized users. Doing it manually is
prohibitively costly and slow. The automatic system under construction does
wide-range intelligent information retrieval and data mining by combining
the innovative ontological-semantic approach with the currently existing
lower-end methodologies and applies it all to the problem of sanitization,
declassification, downgrading, and security control of both textual and data
information. The truly innovative feature of the system is the ability to go
beyond keywords and their intricate Boolean combinatorics to the actual
meaning of the text, thus approximating human performance and overcoming the
unacceptably low ceiling of 80-85% accuracy in the available COTS and GOTS
products, all based on keywords and (rarely) limited syntactic and
statistical information aspect. The system is also be highly customizable
for individual conceptual domains or their combinations.

About the Speaker

Victor Raskin is Professor of Linguistics and Coordinator, Natural Language Processing Laboratory, at Purdue. He is also Charter Member, Internal Advisory Board, CERIAS. Together with Professor Mikhail Atallah, he has pioneered research on the interface of natural language processing and information security. For more on Professor Raskin\'s accomplishments check out his entire biography and CV

here.

Footer

Intranet Site