Allison Parrish: Recent and Selected Work

For more: Smaller recent projects, old projects, my Twitter bots, my GitHub profile, my GitHub gists. My home page has a list of ways to get in touch with me and a current bio.

poster image for Compressed Cinema (with Casey Reas)

Compressed Cinema (with Casey Reas)


Compressed Cinema is a book that collects images from Casey Reas' Untitled Film Stills series. I was commissioned to write a text in response to the images. The texts I wrote are included in the book, interleaved among the images, section by section. Casey generated the images in the book with Generative Adversarial Networks (GANs) trained on films. My texts are "ekphrastic unravelings" of these images: I first wrote laconic, dense descriptions of each image, and then used the mask-filling functionality of a language model (BART) to iteratively "uncompress" those descriptions. More photographs of the book.

poster image for Pocket SP

Pocket SP


During summer 2022, I dug in deep with Game Boy modding and made this: the Game Boy Pocket SP. It's a Game Boy Pocket motherboard that I cut in half and then put into a custom-designed shell with a hinge, a la the Game Boy Advance SP. The build also has a pair of custom-designed flex PCBs to make routing signals between the two halves of the board easier. Along the way I taught myself CAD (with FreeCAD), PCB design (with KiCad) and 3D printing. What does this have to do with the rest of my practice as a computer poet? No idea! But I had fun. As featured on Retro Renew!

poster image for Wendit Tnce Inf

Wendit Tnce Inf


Wendit Tnce Inf is a book of asemic prose poems that I wrote, published in a beautiful hand-made letterpress-printed edition by Aleator Press. To generate the poems, I trained a suite of GANs (generative adversarial networks) on bitmap images of random words, and sampled from those GANs to generate new wordforms, pixel-by-pixel. A separate program I wrote then places these words on the page, one after another, line by line, mimicking the process of typesetting English prose. I originally developed the code I used to generate the book as a NaNoGenMo project in 2019.

poster image for Nonsense Laboratory

Nonsense Laboratory


The Nonsense Laboratory is a series of playful tools for manipulating the way words sound and are spelled. Powered by Pincelate, a bespoke code library and machine I developed, the Laboratory invites you to adjust, poke at, mangle, curate, compress and elaborate the sounds of words in written text. The goal of the Nonsense Laboratory is to make it possible to play with spelling, the same way you play a musical instrument or with modeling clay. In addition to undertaking the creative direction and machine learning engineering, I wrote the front-end implementation. Visuals and user experience by Jenny Goldstick and Tim Szetela.

poster image for Ten Thousand Apotropaic Variations

Ten Thousand Apotropaic Variations


Each 200 (of 10,000) Apotropaic Variations is one of a set of 50, containing 200 computer-generated magic words, printed so that they can be easily cut out. This allows each word to be worn on one’s person or to be ingested for protective effect. Each word is a variation on the word "abracadabra," produced by finding this powerful magic word's hidden state, adding a small amount of random noise to it, and re-spelling from the modified vector.

poster image for Reconstructions



Reconstructions is an infinite computer-generated poem whose output conforms to the literary figure of chiasmus. The program samples a sequence of text from a variational autoencoder neural network trained on the Gutenberg Poetry corpus, and pairs that line with a reconstruction of the sequence in reverse order. The lines of poetry thereby produced are semantic and syntactic mirror images of one another, exhibiting a classic chiastic X structure: ABBA. At regular intervals, the system produces a new pair of lines, which are then displayed between the lines of the previous step, creating a chiastic structure in the arrangement not just of words within lines but among the lines themselves. As the program progresses, lines continue to appear in the middle of each pair, leading to an endlessly nested chiastic structure (ABCCBA, ABCDDCBA, ABCDEEDCBA, etc.). Older lines fade as they move away from the center but are never fully erased from the screen. Featured in the ISEA 2020 Art Programme and exhibited in the Computational Poetics show at UC Irvine's Beall Center in 2022. I gave a talk about the project at Iterations 2021.

poster image for Compasses



"Compasses" is a chapbook recently published in Andreas Bülhoff's sync series. It consists of poems produced with the help of a machine learning model I designed as the next step in my exploration of phonetic similarity. This model has two parts: a "speller," which spells words based on how they sound, and a "sounder-out," which sounds out words based on how they're spelled. In the process of sounding out a word, the "sounder-out" produces a fixed-length numerical vector, known as a "hidden state," which is essentially a condensed representation of a word's phonetics. The "speller" can then use the phonetic information contained in this hidden state to produce a plausible spelling of the word. In "Compasses," I used this model to generate new imaginary words that exist in the negative phonetic spaces between the phonetic hidden states corresponding to names of members of well-known quartets. Poems from the Compasses series have been Featured on Verse, BOMB Magazine and NVIDIA's AI Art Gallery. I gave a talk on Compasses at ELO 2020.

poster image for Pincelate



Pincelate is a machine learning model for spelling and sounding out English words, plus a Python module that makes it super simple to do fun and useful things with the model. I led a workshop at PyCon 2020 on how to use Pincelate to analyze the sounds of words, and generate new words with particular phonetic characteristics.

poster image for A Gutenberg Poetry Corpus / Gutenberg, Dammit

A Gutenberg Poetry Corpus / Gutenberg, Dammit


A Gutenberg Poetry Corpus is a collection of three million lines of poetry collected from hundreds of public domain books in Project Gutenberg, along with source code for making your own custom version. The corpus, supplied as a compressed sequence of JSON objects, is especially suited to applications in creative computational poetic text generation. A previous version of this corpus was used as the basis of Articulations. The corpus itself is based on Gutenberg, Dammit, a corpus (along with source code) I made of every plaintext file in Project Gutenberg (up to June 2016), with consistent text encodings and metadata. Here are the slides for a presentation I gave about both projects at ELO 2018.

poster image for Semantic Similarity Chatbot

Semantic Similarity Chatbot


I teach programming, arts and design and a perennial project idea is to make a chatbot that mimics someone or something—a famous author, a historical figure, or even the student's own e-mails or messaging logs. So I made a simple library and tutorial for creating a "chatbot"–based on semantic similarity among lines in source corpus–in order to give those students some sample code to work with and a bit of a head start on concepts and architecture. There's a Google Colab notebook version you can use in your browser without having to download anything. Janelle Shane had some fun with it!

poster image for Articulations



The poems in Articulations are the output of a computer program that extracts linguistic features from over two million lines of public domain poetry, then traces fluid paths between the lines based on their phonetic and syntactic similarities. The book, published in January 2018 by Counterpath Press, is the culmination of an extended period of research in machine learning and phonetic similarity, which also resulted in a paper published in the Experimental AI in Games Workshop in 2017. The source code for some of the techniques used to generate the book are available on GitHub. I gave a talk at Strange Loop 2017 about this project, which includes a short reading of an excerpt from the book.

poster image for Rewordable



In collaboration with Adam Simon and Tim Szetela. Rewordable is a card game played with a deck of commonly occurring letter combinations in the English language. Players arrange these cards to form words of increasing length and complexity, drawing from their own hand or by stealing cards from other players, building onto already constructed words. Rewordable is the result of a computer-aided game design process, using both computational analysis and extensive in-person playtesting. I wrote about the computational design process in a post on Medium. Rewordable was funded on Kickstarter in late 2016, reaching over two hundred percent of its goal, and was published by Clarkson Potter (an imprint of Penguin Random House) in August, 2017. Buy a copy today at your local bookstore or through your favorite online retailer.

poster image for Live creative text synthesis with word embeddings

Live creative text synthesis with word embeddings


Word embeddings are numerical representations of word meanings in a continuous high-dimensional space. Over the course of 2016, I presented and performed poetry produced through signal processing procedures applied to word embeddings, manipulated expressively in real-time. This research was first supported during a research residence at DBRS Innovation Lab, which resulted in a demo app and a comprehensive write-up by Lab staff members. I gave a talk at Alt-AI on some of the relevant techniques and my 2016 !!Con talk also touched on my research in this area. With the latest version of the live interface, I performed live creative text synthesis at We Have Always Been Digital, a performance event presented by the Electronic Literature Organization at the Kitchen (curated by Illya Szilak).

poster image for Pronouncing



Pronouncing is a simple interface for the CMU Pronouncing Dictionary. It's easy to use and has no external dependencies. I made Pronouncing because I wanted to be able to use the CMU Pronouncing Dictionary in my projects (and teach other people how to use it) without having to install the grand behemoth that is NLTK. I spoke about this library at PyGotham 2015 in a talk entitledd Time to Rhyme: the CMU Dictionary and You. Also available in Javascript flavor.

poster image for The Ephemerides

The Ephemerides


The Ephemerides is a Twitter and Tumblr bot that takes a randomly selected image from NASA’s OPUS database—a repository of data from outer planet probes like Voyager, Cassini and Galileo—and posts it to Twitter, accompanied by a computer-generated poem. The idea behind the bot was to address the similarity between space probes and generative poetry procedures and ask what would poetry written by a space probe look like? The text of the bot comes from two sources: Astrology by Sepharial and The Ocean And Its Wonders by R. M. Ballantyne, both available from Project Gutenberg. The first text contains references to the planets and their movements, and how those movements can be interpreted; the second text is about the open sea, water, ice and lengthy, often one-way voyages into the unknown. The Ephemerides was the topic of an interview I did with Lenny Letter in October 2016.

poster image for Our Arrival

Our Arrival


Our Arrival is a procedurally generated novel by Allison Parrish that recounts the daily events of an expedition through fantastical places that do not exist. The novel’s primary source text is a database of over 5700 sentences drawn from a curated subset of the Project Gutenberg corpus. Each sentence was selected based on semantic and syntactic criteria, namely: the sentence must not have any nouns that refer to human beings; the sentence must have as its subject some kind of natural object or phenomena; the sentence must not have a pronoun as its subject; and the sentence must be in the past tense. The resulting list of sentences appear to be assertions about the natural world. A number of different procedures produce the sentences that comprise the text, including automated grammatical constituent swapping (subject noun phrases are switched out between two random sentences), random topic elaboration (sentences using a pronoun to refer back to and elaborate on the subject of a previous sentence) and random grammar-generated expressions of awareness and affect on the part of the novel's two main characters (I and you). The novel was featured in chapbook form at AND Festival's 2016 Art of Bots exhibit, and excerpts, interleaved with an expository essay, were published in Ninth Letter. Our Arrival was also included in "SAFE", an online exhibition curated by Christpher Clary.

poster image for smiling face withface

smiling face withface


smiling face withface is a Tumblr bot I made that generates and posts glitchy versions of emoji, based on the open-source SVG files released as part of Twitter’s twemoji project. A Python program selects an emoji SVG file at random, adjusts the markup and numbers in the SVG file, and (optionally) recombines the paths in the selected SVG with paths from other emoji SVG files. The results are posted to Tumblr. I wrote more here about my motivations for making the bot. Featured in physical form at AND Festival's "Art of Bots" (great photo of the print here).

poster image for International Jetpack Conference

International Jetpack Conference


Made in collaboration with Rob Dubbin, International Jetpack Conference is a retro adventure game, created for MS-DOS (with ZZT, the once-popular game creation software). The game made an appearance on PC Gamer and was an Honorable Mention for IGF 2015's Nuovo Award.

poster image for A Travel Guide

A Travel Guide


A Travel Guide is a web-based, location-based, mobile-centric application for randomly creating short, poetic texts in the style of the travel guide. A Travel Guide has as its goal to give its visitors an alternate reading of place, through the serendipitous juxtaposition of their current location with evocative procedural text. The guides, of course, are not traditionally "accurate." You may need to try harder than usual to apply the information contained in these guides to the locations in question. A Travel Guide is a 2014 commission of New Radio and Performing Arts, Inc. for its website, and was a featured work in the Electronic Literature Organization's 2016 Festival Exhibit.

poster image for I Waded in Clear Water

I Waded in Clear Water


My entry for [National Novel Generation Month 2014] was this book, generated primarily from the text of Gustavus Hindman Miller's Ten Thousand Dreams, Interpreted (1901). Interpretations from the book were procedurally rearranged to be in the first person, and ordered by automated sentiment ranking. Footnotes were generated from semantic assertions in ConceptNet. I Waded in Clear Water was presented in January 2015 at WordHack, NYC's monthly evening of performances and talks exploring the intersection of language and technology

poster image for Lexcavator



Lexcavator is a retro-styled arcade/word game for Mac/PC/Linux that combines platformer controls with word game puzzles. The goal is to advance your character (@) deeper and deeper into a well of letters by clearing words from the board. There are several game modes: in Arcade Mode, the screen scrolls up automatically, and you need to descend quickly to keep the pace; the scrolling speed increases gradually as you progress. In Time Trial mode, you have a set amount of time to descend as deep as possible. Quest Mode pits you against dozens of puzzle-like challenges (e.g., find three seven letter words in a row). The game keeps detailed records of your play, including a list of the best words you've made across all modes. A minimal, anonymous global leaderboard shows players how their scores stack up, and alerts them when they've found a word no one else has found before. Programmed with I also wrote the soundtrack, which is available on Bandcamp!

poster image for Autonomous Parapoetic Device

Autonomous Parapoetic Device


The Autonomous Parapoetic Device is a self-contained and portable machine that generates poetry. Constantly creating new sequences of words, lines, and stanzas, the APxD promises serendipitous encounters between aleatoric (but affective) text and our experience of physical space. The text that the device generates is ephemeral: it remains on screen for only a small time, and then is replaced with a new (although algorithmically similar) text. No two interactions with the APxD are the same, leading to an endless variety of possible interpretations (is it descriptive? oracular? nonsensical?). The device consists of an Atmel microcontroller connected to a 20×4 LCD screen, and runs for hours on two AA batteries. Conveniently packaged for use in the home or on the road. Featured in Codings at the Pace Digital Gallery (curated by Nick Montfort) and in Neural Magazine.

poster image for New Interfaces for Textual Expression

New Interfaces for Textual Expression


New Interfaces for Textual Expression is a series of devices intended to create and manipulate text. Analogous to contemporary work in the field of New Interfaces for Musical Expression (NIME), New Interfaces for Textual Expression are intuitive but not literal: they map gestures not to characters (as with conventional writing devices, such as the keyboard and the pen), but to broader manipulations of language and layout. The devices suggest new syntaxes for composing, reading, and performing text. Originally consisting of my thesis paper and project at ITP, New Interfaces for Textual Expression continues to be a research methodology for me and an avenue for producing new creative work. Read my thesis paper here.

poster image for @everyword



@everyword is a Twitter bot that I made in late 2007. Its mission was to tweet every word in the English language, in alphabetical order, one word at a time, every thirty minutes. It completed its task in early June 2014. Along the way, it gathered over one hundred thousand followers. Featured in Gawker, the Washington Post, the Paris Review, and more. A collection of the bot's tweets, along with an introduction explaining my process and motivations, is available from Instar Books.