Condors’ Polyphony and Jawed Water-lines Catapulted Out: Gnoetry and its Place in Text Processing’s History

Chris Funkhouser and Andrew Klobucar situate the poetry anthologized in the recent collection "Gnoetry Daily, Vol. 1: a collection of poetry written interactively with computers" within a long genealogy of computer aided writing in order to illustrate how that genealogy continues to be both aesthetically generative and socially significant.

Editor's note: The following essay mixes the voices of its authors, Christopher Funkhouser and Andrew Klobucar.

Offering an example of how one text can be manufactured from another, we - in this case Funkhouser - here use Jackson Mac Low’s expansion method, “reading through” an early draft of our paper, using the letters of its title as “input.” The results partly capture the paper’s essence, limned to become a poem about a pedigree we discuss below. Engaging in this process relates to the subject of our talk, and helps us identify a particular lineage of algorithmic writing (while simultaneously learning something about the type of words we use when we compose a conference paper).

Gnoetry is the name of a software application conceived by Eric Elshtain and Jon Trowbridge, built by Trowbridge, which also became the name of a small collective of creative researchers make processed poems with computers. As chronicled by the “Beard of Bees” website, authors involved with this, “on-going experiment in human/computer collaborative poetry composition,” have collectively engaged with digital textual processing for more than a decade.See http://beardofbees.com/gnoetry.html and also http://gnoetrydaily.wordpress.com/. In late 2011 their first anthology, Gnoetry Daily, Vol. 1: a collection of poetry written interactively with computers, containing poems created with programs named Gnoetry (and eGnoetry), charNG, Infinite Monkeys, ePoGeeS, welatanschauung, JanusNode and other software was released, along with accompanying commentary by their authors: Eric Elshtain, eRoGK7 (Eric Goddard-Scovel), Matthew (Lafferty), edde addad, nathanielksmith, and DaveTolkacz.Note: while a number of these programs function, or are available for download online, Gnoetry itself is challenging to run for anyone not using Linux.  This collection represents the most recent significant collection of text-based machine modulated poetry.

Let’s briefly compare Gnoetry Daily to the first major print anthology of computer poetry, R.W. Bailey’s 1973 Computer Poems. Providing clear evidence of progression in digitally processed poetry, the Gnoetry anthology captures the attributes of a unified research group inclined to divulge compositional processes (in a way that few authors in Bailey’s anthology partake). The anthology edited by Bailey marks an important cultural transition from a more traditional modernist perspective on poetic form and procedure to the unique relationship to language exemplified in programmable poetics, providing one of the earliest substantial introductions to the use of computer programming in the literary arts. His preface invokes a kind of neo-Futurist position describing computer poetry as “warfare.” Language is not only being re-considered in terms of computational programming, it is supposedly liberated. At the same time, specific poetry genres (or sub-genres) associated with modernism do more than merely transition to the screen; they are actually being fulfilled as methods. “Concrete” poems appear even more concrete when produced with a computer–imagism is more imagistic, sound poetry becomes more aurally sophisticated. Bailey is impressed by the evident materiality and object-ness of language when processed by a computer program. But where such attributes allow the poets/programmers of Gnoetry to focus on the process of text generation as an important poetic aim in itself, in Bailey, the works remain part of specific meta-historical narrative beginning with Mallarmé’s dice throws as model for the use randomizing algorithms.   

Thus it is relevant that Funkhouser re-processes Bailey’s preface as Gnoetry Daily’s “Foreword,” as a critical response to the modernist ethos of progress evident throughout the anthology.The processing was done with using the GTR Language Workbench developed by Klobucar and David Ayre. With Funkhouser’s method, Bailey’s preface becomes more of a De-face, transcreated by filters, in which “Computer poetry is warfare carried out by other means” becomes, “Condors’ polyphony and jawed water-lines catapulted out.” This approach also recalls addad’s advice: “[e]verytime you encounter a research artifact (algorithm, toolkit, corpus, result, …), ask yourself how it might be used to generated poetry” (39). Work presented in Gnoetry Daily ostensibly retains three of four main historical tendencies identified by Bailey: “poetry of sound in verbal orchestrations,” “imagistic poetry in the juxtaposition of the unfamiliar,” and “haiku,” (ii) but also interrogates any pretense to meta-historical continuity.

The eponymous software program that brought Gnoetry together synthesizes language based on its analysis of existing texts, thus mimicking the “statistical properties” of its input texts; i.e., users filter language by interactively applying constraints. Poems in Gnoetry Daily are not generated through programming but rather delivered through processing known artifacts. What occurs is either a synthesis of multiple texts, fused together according to programmatic stipulation, or a single text whose language shifts as a result of computer processes.

These operations recall those conducted within TRAVESTY, an analytical innovation in combinatoric writing produced in the 1980s by Hugh Kenner and Joseph O’Rourke. In TRAVESTY—itself a collaboration between Kenner and O’Rourke—users are prompted to insert input text, and to set the desired amount of output and the size of the pattern length (up to nine characters in the original version of the program); the program itself supplies no dictionary or database. TRAVESTY analyzes a text sample by identifying successive patterns of letters and spaces and making a “frequency table” for each character group in a document’s source text (Hartman 55).The successive patterns of letters and spaces are called “character groups” by Kenner and O’Rourke.  This connection is not only aesthetic (e.g., that the output of Gnoetry and similar programs can resemble those of TRAVESTY), but also social because the two programs influenced small communities of experimental digital writers whose practice benefited and expanded due to camaraderie resulting from use of common programs; they have become mileposts for the discipline. TRAVESTY scrambles (or permutes) text by replacing each character group in the text with another (of the same size) located elsewhere in the source. Works by other authors had been used as source texts for databases in the past (Theo Lutz, Nanni Balestrini, and others), but TRAVESTY’s approach to creating a digital poem involves a “manipulation” rather than a “generation” of text (Hartman 95). TRAVESTY particularly highlights human input’s imperative role in choosing a computer poem’s source or database.  In TRAVESTY words or phrases are not recycled, but the combination or patterns of letters in the words themselves and spaces between words become the basis for the program’s output. Though initially statistical in character, these objective, analytical qualities may be later subject to personalization (or not) by an author exercising editorial prerogative.For more detailed discussions regarding TRAVESTY, see Funkhouser, Hartman, Kenner and O’Rourke. 

This “literary lineage” evokes a very important, perhaps even profound, relationship to information as both a cultural construct and a particular way or method of using language: This latter sense of information as a formal method for organizing data can be seen below, where an “Order 6” Travesty is applied to an early draft of our essay, producing meaning simply by reorganizing language use in a particular pattern:

d's poems in Digital Poetry is the stone.

Mist traits own object (1995, 181).

•Definite Monkeys, ePoGeeS to generators. Each generating verbal or forms of the program, and political comparison”, “Research artists – inform

This is productive words.

Although abstract in places, a certain sense can still be gleaned from the text thanks to the patterned re-processing of the original work as new information, formulated, in part, by inputting and re-organizing new variables. In a literary context, such computational practices obviously become poetic. 

Part of TRAVESTY’s significance derives from the continuum of works it inspired. Charles O. Hartman and Mac Low began using TRAVESTY and its derivatives. Hartman believed TRAVESTY examined “the relation between the original and its transformation and deduce[d] various things about the language of the original,” using it to construct a long poem entitled “Monologues of Soul and Body” (Hartman 54). Subsequently he explored permutation and combinatoric possibilities by creating DIASTEXT in the late 1980s. Mac Low favored Hartman’s program and used it to compose several poems and books. Mac Low had created Virginia Woolf Poems using a “diastic” method in 1963, whereby a phrase (or even a word) from a text is chosen, and then words in a source text that share the same verbal or letter patterns are extracted and used to create a new poetic work. Transforming Mac Low’s arbitrary method into a program was not difficult because the process itself is algorithmic and does not involve random elements. Mac Low also began working earnestly with Jim Rosenberg’s DIASTEXT and DIASTEX4 (which allows the user to choose and employ a separate index instead of using the whole source text as the index), along with TRAVESTY, in 1989. These programs profoundly influenced his title 42 Merzgedichte in Memoriam Kurt Schwitters (1994). Hartman’s program mechanically accomplished—with some variation and advancement—the procedural work that Mac Low had practiced for many years (which involved a significant degree of systematic editing and author intervention). TRAVESTY’s concepts and programmatic effects also directly influenced Michael Dickman’s TextMangler 1.2 (1996), in which readers load a text file into the program and TextMangler rearranges them using a “Markov table,” producing randomness so that at any given moment the text’s future is independent of its past—i.e., one piece of information in a text bears no influence on another. TextMangler uses THINK Pascal source code to “display the result of ‘mangling’ a file or calculating a character frequency table” (Dickman). As we will see below, in the discussion of Gnoetry, this table (or “chain”) can be useful for the reconfiguration of poems, although though the raw data it produces may not appeal to popular aesthetics, especially when the character frequency setting is low. Markov chains tend to be more compelling when applied to longer documents, and do not in themselves create line breaks. Like TRAVESTY, TextMangler succeeds in disfiguring text—essentially destroys a poem in the process of making one.

The sub-title for the initial publication of TRAVESTY code (BYTE magazine, 1984) was “[n]onsense imitation can be disconcertingly recognizable” (129). Semantic continuities between successive programmed texts were somewhat “disconcerting” to TRAVESTY’s authors, for they implied echoes of prior verbal biases in all communication); yet for Gnoetry such developments evoke a communal level of engagement. As processed information, programmed text literally re-constitutes the writing process as innately interactive, where its value derives from its ongoing distribution through a network. This instrumental or functional quality, in turn, implies a networked authorship or discursive community.

With Gnoetry, the network (and/or the discursive community behind it) remains an important source of cultural continuity. Linguistic modeling tools like n-gram generators, where aspects of semantic consistency are repressed in favor of variation, randomness and uncertainty, are preferred. Remnants of a more formal semantic framework derive not from a single lexicographic system, but from various source texts themselves, rearranged via Markov chaining or n-grams, which provide, “a type of probabilistic language model for predicting the next item in such a sequence in the form of a (n-1)–order Markov model” (Wikipedia). Importantly, however, while at the core of many of the anthology’s poems, use of Markov chains and n-grams are bookended by supplemental, objective and subjective crafting by the poet. Choosing the database and manipulating output present a more inviting aesthetic atmosphere; readers encounter something crafted poem than an awkward, unshapely stream of letters. 

In Gnoetry Daily, Vol. 1, eRoGK7 explains how his (and by extension the collective’s) approach disrupts “the author-ego complex” and makes “the activity of writing into something less self-involved than more ‘traditional’ ways of writing” (11). His “Why is there a prison” employs a “chatbot,” or a “type of conversational agent” programmed to “simulate an intelligent conversation with one or more human users via auditory or textual methods.”http://en.wikipedia.org/wiki/Chatbot eRoGK7 instructs his device, named “gertbot,” with “lesson plans” combining lines from Gertrude Stein’s writing with lines made by Gnoetry with Stein’s work as source, along with his own “Stein-influenced dialogues;” the aim is to engineer a poetic voice, that “speaks” using syntax and diction similar to Stein’s.http://gnoetrydaily.wordpress.com/2010/11/30/three-gertbot-lesson-plans/. Among the developments that has arisen with text generators involves programming into an application grammatical and stylistic traits of a known writer, as seen in Jim Carpenter’s Erika and Millie Niss’s The Electronic Muse. Stein’s unique form of diction and verbal construction has also been mimicked previously by Mac Low’s “PFR-3 Poems” and by Erika.See Prehistoric Digital Poetry (45) and New Directions in Digital Poetry (124) for discussion of these examples  In eRoGK7’s Gnoetry Daily text, because the first line declares, “Why is there a prison,” the accumulating lines seem to elucidate, without interference and with poetic charge, the confines humanity builds to legislate its people. eRoGK7’s direct and thorough implementation of Stein as verbal and conceptual influence, not to mention his own rhythmic understanding and interpretation of her sense of verse, takes the results to a new level. The poem embraces, but is also uncertain of, fear: 

I believe in terror.

This is not the same thing that is all.

There is no arrangement.

It is very likely to me. (13)

In the end, the lyrical back and forth, its multidimensional address from a supposed personal (“I” formation), almost seems stoical as its lyricism levels the perils of the identified subject into something not insurmountable. 

Elshtain inserts his own micro-anthology-within-an-anthology in Gnoetry Daily Vol.1, featuring works by “my fellow end users and programmers” (1). This gesture of generosity is in keeping—if not indicative—of the collective spirit of Gnoetry, and his selection contains one of most stellar examples of remix poetry on record, eRoGK7’s “FREE GRASS.” Made with Gnoetry, this series of ten haikus integrates Lawrence Lessing’s Free Culture and Whitman’s Leaves of Grass. Instead of one author, or many, we are explicitly given two (not to mention eRoGK7 and Gnoetry). Given Whitman’s sensibilities and style it is relatively easy to dissect these classically organized pieces. Lessig, who proffers liberal ideas about copyright and trademark with regard to technological application, is not at all known as being humorous—nor is this a trait ostensibly reflected in Whitman’s work. However, the way their voices fuse is comedic, as when he refers to the Marx Brothers within the context of a typical Whitmanic brotherly invocation:

O my brothers and 

Warner Brothers, and the Marx

Brothers and sisters. (19)

These haiku arrive at poignant conclusions, as in:

On average, we

must be a violation

of democracy. (19)

Through them he perhaps even emulates what Whitman might have ecstatically written about today’s technological society:

I love the world. No

doubt I have the Internet.

Sparkles from the world! (19)

Beyond the point that is made in many mashups—that different works of art can be productively synthesized— eRoGK7 fuses his sources two and make something genuinely heretofore unimaginable hearing from either author’s voice. Messages projected use external texts to address the introspective concerns of a virtual author, as in the passage: 

My life: Some of what

was a human, with links to

pictures and writings. (20)

At the series’ conclusion, eRoGK7 offers a sober self-evaluation that indicates the intention of his actions, at least from one perspective:

This is piracy,

to exchange content on the

surface of poems.

No mistake can be made about this point: in these poems is a type of poetic piracy. Fortuitously, words are not only stolen from the original author but are returned as reshaped, provocative entities. 

addad’s “! #0p3” (“I Hope”) embraces “leetspeak,” blending—alongside instances of codework and mezangelle—SMS, alphanumeric type, and other symbols in place of conventional language; words are rendered as graphi-phonetical representations.As noted in netlingo, leetspeak was originally used by hackers and gamers but is moving into the Internet mainstream.   addad ups the ante with regard to transcreation. Beyond the procedural combination of original texts by Goethe, Prince, Joy Division, and an NSF Grant Proposal Guide, his notes acknowledge, the work is “post-processed with leet charfont and codework insertion mappings using JanusNode” (40). Did the original output compel addad to re-process the text? On one register the poem’s opening lines, which ask the question, “How deeply am I too conscious of the primary means/of hate identified below. Future days have power”, seem to declare necessity for altered forms of speech (40). In this example, a contemporary lexicon addresses an old issue. Further, “! #0p3” points back to corporeal and romantic matters as rueful, and untangling oblique speech takes effort: 

Cur53d, 70 y0ur m4n 0f :f.!r3[w0rk5] y0ur 80dy,

7h47

10v3,

70 :c.ru[c!fy]5h m3 10v3r. (40)

In addition to Mez-esque splits, decoding the letters obstructs the flow of the task, as readers cobble lines and stanzas together bit-by-bit. The concluding lines, “I hope, to wander the/crowd,” simultaneously recall and counter the traditional Romanticist Wordsworthian image of the poet wandering lonely as a cloud. Here, the poet as programmer indicates an explicit aesthetic and political need to be part of a larger contemporary whole, especially since leet is not typical compositional rhetoric in literature. The addition of this component bestows, beyond poetic feeling and sensibility, a type of present-day semiotic relevance.

Unsurprisingly, addad uses a variety of programmatic methods, and a range of seed texts, to achieve myriad results. The more lyrically-oriented language of “rock out,” made with charNG (with 4-character n-gram Markov chaining) may be discerned as unsurprising because its seed text is a hip-hop song by Redman, and the poem reaches its most interesting heights are its first two (of four) brief stanzas, the language of which, in spite of its echoic repetition (rooted in the original lyric), do not recall anything I have ever heard in a hip-hop song, as in the opening stanza:

Shake

       testing

            your

witch

               unbuck

  out (35).

Here the poem seems to pronounce orders, to actively explore mischief and proceed without fetters. The condensation of language—of finding a way to say a lot with few words, as well as using fragments of words (e.g., “unbuck” above, “pract” below) which retain expressive value—is especially notable and enable refreshing tone in the poem. “rock out”’s second stanza begins with a word that combines fragments of Redman’s words “Proclimator” and “cockeyed," as follows:

  Prockeyes for

       pract

thing

       out (35).

Although in the end parameters addad establishes, as well as any editing he does to the poem in miniature, mirrors the vigorous inclinations pronounced in its seed text, doing so while shaping alternative orientations. Rather than present a pro-sex anthem, as does Redman, we read in “rock out” potentially both a violent poem (depending on how “kill” is interpreted in the lines “Shake to/kill/Right/up”) or as something containing more innocent and youthful energies (“Palm in/Palm”) with its sense of what it is to “rock/out” (35).

Programs used by these authors analyze minute details within language elements as opposed to cultivating meaning through more formulaic structures. In this legacy, Gnoetry represents advancement not because they have the best machines, but by putting an extreme amount of care into focus, process, and production. In statistical processing, combined with output manipulation, less randomness emerges, with an internalized force that definitely emerges in their synthetic approach. 

addad’s eePoGeS or Poetry Generation Sketchbook project also acknowledges and enframes information’s overall socio-cultural worth. eePoGeS provides the capacity to process source texts in legacy fashion by isolating and making manifest specific patterns derived from semantic and syntactic structure of source texts. The construction process involves several notational models and discursive devices, some based upon N-Gram formulations and other on phonemic patterns or frameworks. In one option, the program presents users with several ready-made source texts culled from Shakespeare’s sonnets. In addition, and even more significantly, the user is given the choice of submitting his or her own specific source text, which can accordingly be parsed and analyzed using the word or type-based bigram model, along with the program’s phonetic and rhyming tools. The result is a more dynamic structure, built from probability, but less predictably:

Computer poetry is reflected with a computer poetry

is reflected with a computer poetry in this power

but is reflected with a computer poetry in the surface

of the poet-programmer finds this struggle, our infinite.For this example, Bailey’s preface was used again, where the output was set in the following manner: Number of lines is set to 4, aabb Rhyme Scheme with Enjambment, trailing Newlines, Stochoastic Beam Search of a population size of 6 and number of iterations = 5. Many of these settings are default. 

Explorative works produced through connections among writers around all these programs, as in the attentive focus on hypermedia in Alire, as well as concerted initiatives on the network, such as remixworx and Flarf, prove communities are productive for authors working with computers, just as they were useful to writers in past generations (e.g., Black Mountain School, Beat Generation, DADA, Oulipo). Maybe not just useful, but necessary… 

Process-based and appropriation-based “mash-up” writings can no longer be considered something new, but these authors gain vitality due to their use of relevant source materials. Works represented in Gnoetry Daily, do not simply restate given texts but in their juxtapositions enable the best type of remix in which new sensibilities emerge as poems. Behind the technologies, behind the software and programming, important practices of social engagement and interaction also take place. Disparate texts are threaded together, and whether these new individual fragments retain or excise completely their original contexts, spirited, refreshing artifacts still emerge. The work discussed here calls to mind many valuable social traits - in particular an important sense of generosity whereby processes of sharing and exchange are considered central to all writing practices. We hope such generosity will be passed along and embraced in the works of future researchers.

Works Cited

addad, edde, eePoGeS (Poetry Generation Sketchbook). Computer program. 2011.
---, ed. Gnoetry Daily, Vol. 1. http://gnoetrydaily.files.wordpress.com/2011/06/gnoetrydaily-volume1.pdf, 2011.
Ayre, David and Andrew Klobucar. GTR Language Workbench. Computer program, http://web.njit.edu/~newrev/3.0//workbench/Workbench.html , 2007.
Bailey, Richard W., ed. Computer Poems. Drummond Island, MI: Potagannissing P, 1973.
“Chatbot,” http://en.wikipedia.org/wiki/Chatbot.
Dickman, Michael. TextMangler 1.2. Computer program. 1996.
Elshtain, Eric and Jon Trowbridge. Gnoetry 0.2. Computer Program, 2007.
Funkhouser, C.T., Prehistoric Digital Poetry: An Archaeology of Forms, 1959-1995. Tuscaloosa: U Alabama P, 2007.
Hartman, Charles O. Virtual Muse: Experiments in Computer Poetry . Hanover, NH: Wesleyan University P, 1996.
Kenner, Hugh, and Joseph O’Rourke. TRAVESTY. Computer program. 1984.
---. “A TRAVESTY Generator for Micros.” Byte 9.12 (Nov. 1984): 129–131, 449–469.
“Leetspeak,” http://www.netlingo.com/word/leetspeak.php.
Mac Low, Jackson. 42 Merzgedichte in Memoriam Kurt Schwitters. Barrytown, NY: Station Hill P, 1994.
“N-gram,” http://en.wikipedia.org/wiki/N-gram. 

Cite this Essay: