Work Description

From the author's foreword: "The idea of programming a computer to generate poetry is not new. Originally, this idea took the form of a program to generate grammatical strings, or sentences indistinguishable from lines of poetry if semantic incompatibility were ignored. A number of simple attempts to produce verse were made, but the results had very little theoretical interest and no poetic value. More than one critic cast discredit on the entire notion of computer poetry. Yet a serious rationale can be found for such a project: the discovery of the nature of poetic constraints on the natural language base and the discovery of the relative complexity of each component. The purpose of RETURNER is more specific. By analyzing a particular poem, “Return” (1968) by Alberta T. Turner, and devising a suitable algorithm based on its vocabulary and structure, I expected to produce stanzas of a similar kind which could be arranged into the form of a poem like the original. I also expected that this task would gradually require the solution of a great variety of interesting problems in language synthesis. Work on the program was begun in September 1969 and finished on December 15, 1970. The earlier versions were run on the IBM 360/40 in the Cleveland State University Computer Laboratory. The later versions were run on the larger 360/50 which replaced it. The RETURNER program itself consists of 284 SNOBOL 4 instructions, which can generate one hundred short stanzas on the 360/50 in about ten minutes, or one stanza in about six seconds. Despite a tiresome lack of variety, both in syntax and in vocabulary, these synthetic stanzas have an unmistakable “poetic” quality and a family relation to the source poem. The poet herself has found this permutation of her artifact of sufficient interest to base a new set of poems on the constructions produced by RETURNER, two of which appear later in this booklet. It would be too simple to ascribe this effect merely to the presence of the same words in “Return” as in RETURNER. Nor do we believe any longer in such a thing as a poetic diction, unless a diction is poetic which collocates crowbars with kittens, dogs and staggering. The unexpected collocations of poets like Dylan Thomas have perhaps oriented us to a different conception of poetic texture or a different sound of the poet’s voice. A consideration of the problems involved in generating computer poetry alerts us to the curious behavior of familiar words in unfamiliar combinations or contexts. We perceive how readily we accept metaphor as an alternative to calling a sentence nonsensical. We tend, that is, always to try to interpret an utterance by making whatever concessions are necessary on the assumption that the writer had something in mind of which the utterance is the sign. Of course, this is inappropriate when the speaker is a computer. The consequence seems to be the demolition of the critical axiom that the poem is sufficient. If we are not to waste our time in vain interpretation we must now ask a new question before beginning an exegesis: Who or what wrote this poem? The problem, however, will not arise in a serious form until computer poetry becomes somewhat better than it has been until now."

Notes

Submitted by the publisher, Aleator Press.

Publisher Note