A

Keith Woodward woodwaka at uwec.edu
Mon Jun 28 12:04:08 CDT 1999


At 09:03 PM 6/27/99 -0400, Ben wrote:

>Okay, now I want to ask a question about a potential strategy for
>determining whether an author has penned a particular piece of writing or
>not, irrespective of whether the author says this is true or not.
>
>I have been told by someone who works with computers that there is an
>algorithm which can be used to determine such a thing.  Given a sample of
>writing (whatever might qualify as a useful example), this algorithm
>supposedly can indicate whether or not another sample of writing was
>composed by the same author.  The person that told me about this said she
>recalled hearing that it had been used to try to settle issues over
>whether Chaucer wrote all of the _Canterbury Tales_ or something.
>
>Does this algorithm exist, and if so does it work?  I need to ask because
>I have no idea and I wouldn't really want to get to know computers well
>enough to find out.

It's called stylometry, a "computer-assisted technique":

"Tha computer can analyze and classify, in whatever ways desired, such
elements of an author's style a word frequency, sentence length and
structure, rhetorical and grammatical forms, metrical and other repetitive
patterns, and special techniques--Chaucer's alliteration, for example, or
Milton's use of names or other elements from a particular source or imagery
of a specific type, astrological, biblical, historical, literary, or
mythological.  The basis of the attempted discrimination or identification
remains the familiar cluster of characteristics that differentiate one
author's style from that of others, or, conversely, reveal stylistic
affinities between two or more authors.  The computer method, however, is
much faster and superhumanly capable of handling complicated relationships.
Perhaps most importantly, it reduces the possibility of impressionistic
conclusions..."

--Richard Altick, et al.  _The Art of Literary Research_ p.96-97

Keith W




More information about the Pynchon-l mailing list