This thesis explores how various computer programs
construct poems and addresses the way several critics
respond to these computer generated texts. Surprisingly,
little attention has heretofore been paid to these programs.
Critics who have given the matter attention usually focus on
only one of the myriad programs available, and more often
than not, such scholarship concludes with a disparagement of
all such projects. My work reexamines computer generated
poetry on a larger scale than previously exists, positing
some conclusions about how these texts affect contemporary
theories of authorship and poetic meaning.
My first chapter explicates the historical debate over the
use and limits of technology in the generation of text,
studying similitudes between certain artistic movements and
computer poetry. This historical background reveals that
the concept of mechanically generated text is nothing new.
My second chapter delineates how the two main families of
computer poetry programs actually create these texts.
Computer programs combine existing input text, aleatory
functions, and semantic catalogues, which provides insight
into how humans both create and interact with these
programs. At the same time, this study illustrates the
difficulty in defining the level of intention and influence
by individuals on the textual product, and therefore these
texts challenge our traditional notions of authorship and
the value of poetry. My third and final chapter argues that
contemporary literary theory and poetics creates the
conditions under which computer generated poetry can pose as
a human product. The success of these programs to deceive
readers about the origins of the text becomes clearer with
the results of a survey I conducted in which the respondents
were fooled by the machine more often than not. This
possibility of machine-created text masquerading as human
art threatens many critics, who quickly dismiss the process
and its results as non-poetic, but I conclude that since the
computer complicates foreknowledge of origin in some
contemporary poetic forms, this intrusion by the machine
prompts us to reconsider how we traditionally value and
interpret poetry.
See above.
"Formulaic poetry generating programs produce texts
influenced by two individuals : the programmer and the
operator. One could argue that they are one in the same,
since by inputting data such as subject and gender, the
operator enters into the role of programmer and "finishes"
the instruction set. It would follow that in such a case,
the label "programmer" now applies to a role and not to a
specific individual. Much to the possible disappointment of
the Bill Chamberlains and Chris Westburys of the programming
world, authorship now disintegrates into a true author
"function," not applicable to identifiable individuals. Yet
somehow this creates a nagging sense of inaccuracy precisely
because of the type of language computer programmers use."