CHARACTERIZING CASSANDRA FOR THE CAMERA
Dr Richard Forsyth, Bristol Stylometry Research Unit, University of the West of England

Abstract: In November 1996 an article appeared in the left-wing magazine Tribune under the pseudonym "Cassandra". This article was heartily critical of Tony Blair in particular and the leadership of the Labour Party in general. According to Tribune's editor it was penned by a "senior Labour MP" who wished to remain anonymous.

A team from BBC2's Newsnight programme were aware of the work of Professor Donald W Foster at Vassar College in uncovering the author of Primary Colors, an anonymous book about Bill Clinton's 1992 presidential campaign. Foster had declared, on stylistic grounds, that the author of Primary Colors must be Joe Klein, a Newsweek journalist. After publicly denying the attribution on TV, Klein eventually admitted authorship -- only when an employee of the publishers Random House `leaked' to the press proof copies showing corrections in Klein's handwriting! The Newsnight team decided that it would be interesting to attempt a similar coup in the Cassandra case, and called up the Bristol Stylometry Research Unit (BSRU) here at UWE, one of the few places in Britain to specialize in this sort of work.

They provided BSRU with seven text files on disk, labelled A-F, as well as the text of the Cassandra article in electronic form, thus setting up a genuine blind trial. The comparison texts were all articles written by Labour MPs who were suspected of possibly being Cassandra. This presented an unusually challenging authorship attribution problem for several reasons: (1) seven candidates is quite a large field; (2) the disputed text was only 980 words in length; (3) undisputed samples from the candidate authors were short as well, under 600 words in one case; (4) there was a non-negligible chance that the seven candidates might not include the true author.

But Richard Forsyth and David Holmes of BSRU rose to the challenge. How they made their decision to meet a strict time deadline; how they tried to explain it for the media; and how it came over on TV will be revealed in this talk -- as will the probable identity of Cassandra her(?)self.

This seminar was held at the Department of Computer Science, Royal Holloway, University of London on 20 May 1997.