A Long History of Generated Poetics: cutups from Dickinson to Melitzah

This is a transcription of a talk originally given at Wordhack, a monthly event at Babycastles exploring the intersection of language and technology.

When talking about generative text, there is often a predilection to start with its histories. This is perhaps an impulse towards humbleness, an important recognition that although computers have radically shifted the field, they certainly did not invent it (and are only tools, after all).

Generative poetics has the obvious roots in Dada, in the cut-ups movement, in found poetry and the modern-then-post-modernists who (although not all working with computers) certainly were operating in an era rife with the promise of automation and a future of intelligent machines. Looking at this work, it is easy to see the seeds of what many of us in this field are pursuing now; work like The Wasteland, or Ulysses, or Burroughs’ many experiments with shuffled text all take advantage of the same thing that contemporary generative poets do; that meaning is mutable, that it builds itself in the mind of the reader as words compound, that we as humans are built-in pattern recognizers and storytellers and that any text need only do half the work; we will fill the gaps, follow the shift, and assemble a one thing to take home with us, even when it is made of many.

But rarely does the history of this space extend farther than the early 1900s, and viewing cut-ups and the fragmentation of language as a fundamentally modernist discovery presupposes one of the basic functionalities of text; that it is modular. Language works precisely because it splits apart and recombines according to set rules; this supposition is the basis of core concepts like the word, the character, the syllabic utterance. Dadaist cutups although a remarkable reflection of their times, were almost certainly not the first instance of pieces of language being reassembled into new forms.

To start, one could perhaps look at the poem-fragments of Emily Dickinson. Although her work was significantly edited after her death in order to appeal to Victorian sensibilities and is only just being published in its original, it was written with little regard to convention. It seems she selected scraps of envelopes, notes, and discarded papers precisely to fit each poem. Dickinson was financially comfortable and need not go to such lengths to preserve paper, nor was she known to jot down poems while out, where notebooks might be absent. These paper fragments were collected to house poetry in a tangible way; the poems follow the shape of the paper, and lines break and shift to fit their borders.

Dickinson numbered each of her poems chronologically, and also numbered are many scraps that seem more like lists or notes than poems, particularly poems in the style of the 19th century. But they are sandwiched between more conventional verse, and together create a body of work that includes within its borders- bare sentence scraps, notes to self, and notably; several recipes. Poems like Kate’s Doughnuts, a piece of what we can only assume is found text, is placed at the exact same scale and importance as all of the Dickinson that is so famous today.

Of course, there are much earlier examples; one could consider Honkadori, a 12th century Japanese poetry intertextual technique in which writers would allude to, or quote, famous historical poems inside of their own text. This allowed a sort of unfolding meaning; someone familiar with the earlier poem would recast the apparent text into a different light, while holding the contemporary framework on top. It allowed poems to exist in two spaces, a sort of layered experience that depended on a recycling mechanic, a borrowing of history as a contextual technique.

Kokinshū 606:
Keeping this longing
Hidden within is what hurts –
With only me to hear my sighs
 ~Ki no Tsurayuki, 9th century
1035:
Another evening’s sighs:
Have I forgotten
This hidden longing
Is mine alone to suffer
As days become months?
 ~Princess Shokushi, 12th century

This fashion could have been an import from China, like many others at the time, or could have risen independently from the rich poetic practice of the Japanese imperial court. Regardless, it bears similarity to pre-imperial Chinese court argument starting as early as 300 BCE. This is a style of text that only reaches its full meaning when listeners draws connective lines from unit to unit of text; standing alone, each unit of argument lacked the power of the structure of a whole. Building on itself, it followed, as William Boltz argues in “the composite nature of early Chinese rhetoric”, a logic of signs that take their power from one another.

Literary Forms of Argument in Early China, Joachim Gentz and Dirk Meyer, page 119

Or perhaps one could consider folk traditions at large, in which historical stories often include within their borders rich local subtexts of family histories, cultural norms, and personal anecdotes that are folded into traditional text- a borrowing of experience that intercuts the everyday lessons of the world into long-standing stories.

And there are likely many many others that have been lost to poor research, destroyed histories, or general negligence.

But I wanted to point specifically to the medieval concept of Melitzah, perhaps one of the clearest example of early cut-ups surviving today. Melitzah (also sometimes called Shibutz, or sometimes classified under the more general Midrash) is a medieval Hebrew literary device in which a mosaic of fragments and phrases from the Torah, rabbinic literature, and the liturgy are fitted together into new meaning, most often in the form of religious argument but also in poems, letters, and personal writing.

Unfortunately, there are very few medieval pieces translated, but here is a later example in an Ukranian letter between brothers from around the 1850s (although primarily medieval, Melitzah came back into fashion in the 1770s during Haskalah, a kind of Jewish enlightenment).

… How could I dare not answer them with some word of instruction to you; for, what could I say and tell and with what can I justify myself!? [1] For how can my heart rouse me to open my mouth and speak words which are not true against the Rule of the Creator? Have I, a slithering worm cursed among people, complained against Divine Providence? And, who is this and what is this that I should express opinion on such matters? It is very true that I have learned how to commit sin against G-d: I have sinned and transgressed before Him. Only in this am I able to justify my actions and to rectify my deeds, only in this: that I am in truth guiltless before the Lord, my G-d and of my fathers. And, I have not done evil against my G-d. Also, if I have spoken out of the bitterness of my soul words which are not true, my feet have not yet stumbled and my steps not slipped [2] for the reason of having read at that time horrible and awful things which all ears would resound upon the hearing of them [3], and people would not believe it if it were told [4].
Then, my soul being cast down [5] and my spirit poured out, bereft of words, I am unable to write down everything that was in my heart, because it was not consoling that I had requested; neither words of succor did my soul desire. For does not my letter demonstrate what lay then upon my spirit? But, now I am not true to myself now that my soul has returned to its tranquility. I shall understand my foolishness and I shall never return to folly [6]. And of you shall I ask that you write me also in the future (more) letters such as these, letters of preaching and knowledge, letters of wisdom and understanding. For your last letter, with its flowery language, captures the heart as do the beautiful words of our holy prophets. Its content is deep and requires exactness, and its utterances are sayings of beauty [7]. Please make known to me knowledge so that I may know. Pray, give me understanding so that I shall understand, and you should know that you will not sow good seeds among thorns and thistles, and that your sagacious words will not become lost into nothingness. I shall tie them around my neck [8] and bind them around my throat [9]. I shall inscribe them upon the tablet of my heart, and together with your memory they shall never depart from me.
Behold your brother, wishing you peace.
Shmu’el.
1) paraphrased from Gen. 44:16
2) from Psalm 73, verse 2
3) phrasing from Samuel I, 3:11; Kings II, 21:12 and Jer. 19:3.
4) from Habbakuk 1:5
5) from Psalm 42, verse 6
6) from Psalm 85, verse 9
7) from Gen. 49: 21
8) from Proverbs 6:21
9) Proverbs 3:3

Translation from Alex P. Korn, Ph.D.

What is so special here is not the conceit of recombinatorial text, nor even recombination with the intent of borrowing authority, but rather that Melitzah is motivated specifically by the presence of holy words. In Melitzah the sentences made of compounded quotations mean what they say, but they also ring with associations and echoes to the original source. Beneath the surface of any particular piece of text is a long history of canonical faith; by this, the new sentiment is influenced, but it also reaches backwards; the historical, divine text takes on a layer of the new. It is a methodology of prayer as much as it is an attempt to express something eternal or empirically true.

Michael Marmur, who has studied the quote-impulse in historical Jewish texts, might have said it best;

Biblical verses are sewn into the text of the poem and come to serve a variety of functions. In point of fact, the lines of distinction between the decorative and the generative aspects of quotation are blurred in the extreme. Just as a quotation beautifies, it also amplifies. The sources become the basis for almost unbridled creativity, and, at its apogee, the art of applying these verses blurs the distinction between the extrinsic and the intrinsic.

Of course, contemporary generative text works in much the same way; fragments are pieced together forming new meanings out of older language. This is quite apparent in Markov-chain type generation (where a text is analyzed for A’s chance of following B), but it also holds true in more complicated techniques like recursive neural networks. Even at their very best, these machines are all about their corpus; without something to learn from, they produce nothing at all.

So much of the work of someone working with these tools is on the side of the input; carefully seeking out the right kind of text, learning how to structure it. What is smart about Melitzah and other historical generative textual practices like it, is that the input- the corpus- is highly specific. One can trust the Torah to have a fairly unified voice, to carry a certain weight, to work in the same way over time. Of course, the writer of these verses was performing the task of our machines by hand; assembling bits and pieces, altering grammar to fit, finding the perfect passage to touch another.

In many ways, these medieval Hebrew scholars were perfectly tuned to this task. After all, they lived in the world of the Melitzah; they were trained to rabbinic literature in the same way that we are trained to emoji or to Wikipedia articles. We understand how they function; we have lived in the content so thoroughly that we know the form of the thing.

This type of training is a possibility for machines. They too can learn how a Wikipedia article is structured, or what sentiments various emoji generally represent, or indeed which pieces of liturgical text flow into each other. They excel at this kind of task- reproducing structure, getting the ‘feel’ of a thing.

So often when I talk about my work to writers unfamiliar with generative techniques, they joke that they will soon be out of a job. It is in jest, but it happens so often I can’t help but feel it comes from a place of genuine panic. I try to explain; sure, my tools can reproduce the right structures but they don’t really ‘get it’, meaningful output or not. Any moment of delight or clarity is on my end- the creative act here is in the reader.

When I say that the creative act is the reader’s, I imply the creator as well as the audience. When working with generative text, it is impossible not to read. One has to look for bodies of text that can function as useful sources for tools; big enough, or concrete enough, or with the right type of repetitive structure; learnable. And then one has to read the output of such machines, refining rules and structures to fix anything that breaks that aura of the space one is looking for. In this, we are not unlike the medieval scholar who studies holy verse to become fluent enough in that space that it becomes building block.

To ask if a machine could really understand the Torah, or for that matter- the importance of an Emily Dickinson recipe or a cutup newspaper poem or Wikipedia or emoji is not a question for now. That is, at this time, still our job; our machines stay tools like any other. What we are learning is what carries through, what we can teach our machines, and what lends them strength. The delight is all ours.