The Scope for a Reader
The Poetry of Text Generators

by Chris Funkhouser

Kissing the Steak: The Poetry of Text Generators

Syntext, developed by Pedro Barbosa and Abílio Cavalheiro in the early 90s (later partially re-versioned on the World Wide Web), is a collection of fifteen computer programs from the 70s, 80s, and 90s that automatically generate various styles of poetry in DOS. Though the texts made by each of the programs are thematically unrelated, through these pioneering works by Barbosa, Nanni Balestrini, Marcel Bénabou, and others, each of the predominant fundamental attributes of text-generators is clearly divulged. Syntext, despite being primitive on the surface, powerfully brings to light the expressive possibilities, versatility, and variation within permutation texts, and provides sufficient evidence upon which a typology of computer poems can be established.

Peter Gendolla writes that virtual installations, such as Text Rain, leave us with a “troubling” question which theoretically applies to all interactive, generated works: “are we reading a poem…or is the poem reading us?” I am not as troubled by this question as I willingly perceive we are reading digital poems, and the poems are reading us. Contemporary, electronic poems may be challenging/intriguing/annoying to read, and also reflect the reader’s identity as a technological being—perhaps in an uncomfortable way. Poems on websites embedded with cookies do literally read the reader—one’s tracks through the tracks of digital poems can be tracked.

Producing digital writing begins with an idea or concept, which is developed by a programmer using computer technology, who shapes, populates, and makes it work. The bulk of the effort involved is on the human side of the equation; the computer carries out orders made by the author. When a program/poem involves randomization, it would seem to give agency to the computer, but it is the author who has called for this and (typically but not always) prepares the database. Much of the work is done by the human programmer; much less is done by the machine, which basically does as told. The programmer/poet creatively conceives the work, establishes the database (which is absolutely crucial), engineers the chosen mechanics of the poem, and pushes it to delivery. Obviously, there are limits to what can be done, and various gauges by which to measure success—more than the critic analyzing a written poem typically contends with.

When we read the digital poem as critics, we are doing what we are trained to do—offer insight into the verbal dynamics of text, applying our own bias, perspective, and knowledge to read the language presented by an author—and then more. While it is both significant and interesting that an artist/programmer has created the work using computer technology, and certainly this is an important aspect of the text to consider, this is really just another potential layer of the text to address. Literary critics should be stern with analysis of the digital poem, while concurrently taking technological dimensions into account when there is something to be said about them (which is often). If the output does not stimulate to any of the senses, we should say so, and explain why—but we should also be sure we understand the intention of the poem, and what it is technologically doing, to begin with. This is especially important for unconventional works. For example, most readers may not aesthetically appreciate the output created by the Hugh Kenner-Joseph O’Rourke program Travesty, but if they know what the program is doing, why and how it makes texts the way it does, they could then develop some sort of appreciation for the text. It is the job of the critic to be sure as many dimensions of the work as possible are seen. Further, the critic, to be responsible, cannot rely on one generated example but rather several, in order to give a more fully rounded account of what the generator program is or isn’t capable of doing. This is impossible to accomplish using a single example of any generated text. Initial impressions are important but need to be fortified by studying and interacting with the work.

There is always intent, but does one meaning in every moment in the process of the production of digital literature? Decidedly not; the critic looks beyond meaning for other possibilities. This predicament is nothing new to literature. Pound himself reflected, embarrassedly, about spewing nonsense in The Cantos. Meaning, through language and lyricism, is not presented in the individual lines of postmodern works (see books by Language Poets, for example). But the sum of the lines does communicate something interpretable, and sends a message to the readers. Perhaps there are people out there creating meaningless works of digital literature, but I somehow doubt it because it is not worth the effort involved!

Readers, explorers of the text, are responsible for a lot in this equation, including the production of meaning. Generated works that involve nonsense may not be welcomed—the reader may or may not be stimulated and will decide whether or not engaging with them is worth the effort. Knowing that a machine is creating output may be a deterrent, finally. Works can be too aesthetically challenging (or challenged), and the “noise” within the encounter of the text may impede access to the reader’s imagination! Fortunately such works will probably make themselves known to the reader almost immediately.

If a program’s database itself is not randomly prepared—that is, if the author has a purpose, theme, etc.—then presumably intentionality is infused. Serious works of digital literature are highly thoughtful, highly engineered communication; it may be wildly unconventional communication, and can be put into context as such. I know of no programs, generators or otherwise, that did not have, at very least, some sort of theoretical idea to communicate. The development of digital forms themselves expands the parameters of what literature is. It also alters our perception of what literature can be—although it is not firmly established in the canon by any means. In the end, digital works may end up being classified as technological literature (or something else)—it is difficult to establish exactly where it fits in given the current academic configuration but develops nonetheless.

The computer is obviously incapable of feeling emotions, but it is capable of projecting or transmitting language, sound, sights infused with pathos. Any emotional transmission is the result of human interjection. The viewer of a certain kind of movie or television program can have an emotional response to content presented via these forms of media, and I do not see why we should be surprised to be moved, disturbed, or even elated by something that emits from a computer screen. Of course this doesn’t frequently happen; it is so rare, in fact, that this is why we are surprised when it does (not because it is happening because of its computerization). The audience can respond as it will, and perhaps as yet shouldn’t have high expectations.

Computer generated poems force us to confront human-machine hybrids and sometimes the combination will be unpleasant, particularly if one expects machine modulated work to compare to written works. But the types of disruptions that occur in digital work are not an absolute characteristic. Whatever the surface and depths the artwork contains can be appreciated, charted, defined, and evaluated by how it is presented. Being not a singular form, interacting with digital literature—which is written for many purposes—is a process of many negotiations. Some works are highly complex and difficult, some are more fluid. Engagement is controlled by the viewer, who has been led to, or gravitated, to the work.

Among the works I have been drawn to are those prepared for a collection of generators titled Syntext, produced by Pedro Barbosa and Abílio Cavalheiro. Works presented in the anthology are purposely composed so that syntax is preserved in randomly generated works through combinatorics, permutation, and slotted methods.

We can read it as we would read literature written by a single author, or can alter our reading standards and look for other aspects in the poetry. While critics such as Aarseth challenge the practice of reading and discussing new works by old standards, one can see the value of applying both old and new critical lenses to the work. New lenses focus on technical, expanded aesthetic issues, and evaluation of processes into the critical mix.

Programs in Syntext incorporate a range of source materials (including texts by other authors). Multiple sources (inputs) lead to more diversified combinations, which lead to more complexities (for both author and viewer), as well as more outcomes. The imposition of dramatic and mysterious elements is enticing, heightened by sophisticated, surprising, verbal juxtapositions in both clearly stated lines and those that resist bearing finite meaning. Successful works sometimes feature an opening “hook” that draws the reader into a speculative internal dialogue that is sustained by characteristics of the programming in the lines that follow. Viewers can be led into the poem in much the same way they are on a page, through speculation and by establishing a ponderous natural setting.

In Barobosa’s “Porto,” the reader confronts impossible circumstances (“the nostalgia of the stone”) among the logical, made possible only through creative reflex, sometimes difficult to interpret or envision. Being precarious, at the edge of the cliff, imaginatively, has always been the work of the poet. Is using the machine to be so cheating?

The absurdity and humor in Barbosa’s “Cityman Story” are refreshing: completely alternative perspectives and meanings, divergent from the original poem, emerge through the randomness of the subsequent lines’ order and shape. The positions of the poem’s phrases and the meaning they produce change in each example, such as when the subject of the poem kisses a steak instead of his wife. Barbosa’s programming design generates a variety of narratives from words originally composed for other purposes. Discursive capabilities of the programs well suit the task of making fragments cohere while enlivening the initial humdrum character of the poem (whose existence is portrayed as typical and narrow). The poem is lively and peripatetic—the program serves unites content and form in an imaginative way. This “cut-up” work adapts the language of the original to produce a series of texts that portray surrealistic (absurd) and humorous characteristics, in which “there are progressive degrees of freedom”. Many distinctive “stories” are told, all using the same language to portray wild deviations from the mundane occurrences found in the source text, driven by Barbosa’s determination to subvert the status quo of his subject. The poems retain a type of narrative while transforming the language into something different, projecting a narrative by something or someone who is seeing the world from an alternative point of view.

Cumulative meaning or understanding is established by the reader, who is challenged to create the circumstance given the authorial framework and loosely directed verbal scheme. In order to believe it is a poem, a text’s content and form must compel the reader. Many if not most generators are thus unconvincing, but when the content of the output presents itself realistically (from any perspective) it is worthy of consideration despite redundancies that occur. One can adopt a Cubist perspective—these permutations are not redundancies but rather different dimensions that emerge from a text.

Dramatic sensibility, surprising imagery, and tormented narrative startle the reader due in part because of their machine modulated condition. The articulations of the programs, written by a person but projected by a computer, may not be taken as seriously as the madness in the lines of Edgar Allan Poe or other poets who included such dramatic features, but they are convincing evidence that emotionally driven content can be projected by the apparatus. Various degrees of humor and irony are also supported in these effectively randomized, unique cyborgian texts. In this process, one poem becomes the foundation of, as Jean-Pierre Balpe describes in his essay "E-Poetry: Time and Language Changes" an “infinite, not eternal” chain of subsequently produced works.

Programs that deliver a range of varied output will reward the viewer, whose effort to produce and consume poems is non-trivial. Anyone looking for a range of unique derivations of texts can enjoy these programs, and could use the output as a starting point for their own expressive articulations. A creative programmer imposes outside (artificial) order and formal structure in computer poems by designing a framework where only certain components are randomly filled by grammatically appropriate words. The author/programmer studies the form, determines an effective linguistic framework, and prepares the database accordingly in order to generate meaningful poetic statements (both in themselves and when juxtaposed with other examples) in which the happenstance merging of elements has the ability to create unusual and paradoxical concurrences.

Meaning can be generated on multiple registers even when strict parameters are imposed on a work if the vocabulary included in the database is versatile. The author/programmer selects words that effectively fuse with others and cohere with each aspect of the verbal equation. Works from Syntext not only demonstrate the flexibility of computer poems, but also clearly establishes that the careful arrangement of elements and negotiations between random factors become the forces that determine the quality of this form of digital poetry. In these examples, motivations beyond reifying basic Dada impulse are present in most works. Generally speaking, the “syn-” prefix affixed to “text” in the title can be taken literally. Works (texts) produced by the programs contained in this title are synthesized and synthetic poems or prose poems that do not spring from a natural, singular source. They do come in to existence somewhere between chaos and order, and deconstruct human language to find new meanings. Literary and cultural routines are subverted by the computer, the programming of which manages to preserve the poetry in a readable and interpretable state.

Technically proficient use of cleverly devised language and capable poetic grammar: seeing these productions is an informative experience, which not only shows that computers can capably co-create poetry but that entire—potentially infinite—readable anthologies of digital literature have been produced, even if two readers would never see the same work. These readable texts are stimulating in several ways and can quickly transform its audience’s mindset, as poetry and literature have done since its condition as an oral form—here with the push of a button instead of turning a page or sitting in the audience. Someone who is interested in reading something that will perhaps disrupt his or her path will especially appreciate these programs.