IP: SHRINK AND IDENTIFY Edupage, February 1, 2002

[David Farber <dave () farber net>]

> Italian researchers at La Sapienza University say they have found
> a novel use for the Gzip file compression program, which could
> help match bits of texts with their true authors. La Sapienza
> associate mathematics professor Emanuele Caglioti said the Gzip
> program, which strips huge text files down to their most basic
> components, can use common identifiers to link text pieces from
> various sources. Gzip and other compression programs boil text
> files down to key elements and write instructions on how to
> reconstruct the complete document. Those instructions, or the
> file's entropy, are nearly the same when from the same author,
> La Sapienza researchers found. They compared 90 pieces from 11
> authors using Gzip data and found the technique correctly
> identified authors of works 93 percent of the time. Caglioti
> said the method has applications for Web searching and has
> already been applied to DNA sequencing. Gzip author Mark Adler,
> however, is circumspect about Caglioti's claims and said that
> further work should be done to test the theory using a much
> broader pool of works and authors.
> (ABC News, 30 January 2002)

For archives see:
http://www.interesting-people.org/archives/interesting-people/