Poetry Machines: Empowering Creative Writers to Design DH Tools

lightning talk
Authorship
  1. 1. Kyle Paul Booten

    Dartmouth College

  2. 2. Katy Ilonka Gero

    Columbia University

Work text
This plain text was ingested for the purpose of full-text search, not to preserve original formatting or readability. For the most complete copy, refer to the original conference program.

Digital Humanists often build tools based on their own understanding of what features users will need, whether intuited or derived from user studies. In the spirit of a key theme of DH2020---Public Digital Humanities---our presentation will consider what happens when we as researchers instead empower people to design their own DH tools according to their (perhaps idiosyncratic) purposes.With our project, "Poetry Machines," we are recruiting individual poets who are not themselves programmers or machine learning practitioners to design their own artificially intelligent tools that will assist or challenge them in the middle of the writing process.This research takes the approach of "participatory design". We have composed a design guide, a "boundary object" providing a way for the poets to design their own poetry machines by combining various basic text processing functions and machine learning techniques into complex interfaces. We will then build these interfaces to their specifications, working with them collaboratively to further refine and develop designs (for instance, by presenting options of how a particular wished-for feature could be implemented).A few possible functions that such machines could perform (the first two drawing from our own research [Booten 2019; Gero and Chilton 2019]): provide stylistic advice, e.g. alert the writer when she has used highly common lexico-semantic patterns or descriptors (e.g. 'glowing' to describe 'moon')replace a word with a synonym that is characteristic of a particular author. whenever the writer pauses for too long, complete the line with a neural networkOur design guide provides the poet with substantial flexibility in terms of how the Poetry Machine is activated (for instance, randomly vs. after every line), how it intervenes (for instance, adding to the poem or displaying information in a sidebar), and how machine learning models are trained (for instance, a neural network trained on Romantic vs. modernist poetry).Our research goals are to determine: 1) whether different sorts of poets (e.g. formalists vs. more experimental writers) make different design choices, 2) whether the machines meet their design expectations, and 3) whether this process provides the poets opportunity to develop critical perspectives on machine learning itself. Having acquired institutional approval to solicit poets for formal research, we are in the process of engaging poets in the participatory design process, and our presentation will offer preliminary findings.This work intervenes in several DH fields:Though we are inspired by key examples of aesthetically-experimental hermeneutic interfaces, our research will be of interest to DH tool designers more broadly, especially those interested in "bespoke" tool."Electronic literature" is part of DH. Programmer-poets often design their own tools for algorithmic co-writing (e.g. Johnston's [2019] neural-network assisted ReRites). Our project aims to make this "e-lit" practice more widely available. This research connects to conversations about using DH to expand algorithmic literacy.

If this content appears in violation of your intellectual property rights, or you see errors or omissions, please reach out to Scott B. Weingart to discuss removing or amending the materials.

Conference Info

In review

ADHO - 2020
"carrefours / intersections"

Hosted at Carleton University, Université d'Ottawa (University of Ottawa)

Ottawa, Ontario, Canada

July 20, 2020 - July 25, 2020

475 works by 1078 authors indexed

Conference cancelled due to coronavirus. Online conference held at https://hcommons.org/groups/dh2020/. Data for this conference were initially prepared and cleaned by May Ning.

Conference website: https://dh2020.adho.org/

References: https://dh2020.adho.org/abstracts/

Series: ADHO (15)

Organizers: ADHO