Project Info


Robot, Found: The Collaborative Human-Machine Poetry Writing Project

Tom Williams | twilliams@mines.edu and Toni Lefton | tlefton@mines.edu

Poets working within the genre of found poetry seek to create poems from fragments of text originating from other sources. Within this framework, it is the poet’s job to select and rearrange these fragments of text to reveal new insights and commentaries relating to the technical, social, or cultural context in which the original text was generated.

Advances in natural language generation allow semantically meaningful fragments of text to be generated on any given topic. Within the computational creativity community, researchers have been actively developing methods for computational poetry generation, in which entire poems are generated by the computer based on some initial input.

This project explores a novel combination of these two bodies of work, by (1) developing new techniques for computational poetry generation, and (2) using those techniques to produce text fragments that can be used for the purposes of found poetry generation. We believe that this approach may provide a powerful new technique for the writing of poetry on themes relating to artificial intelligence, automation, social media, and other technology-related themes.

This year long project is expected to unfold in two stages. In the first stage the student will explore existing frameworks for computational poetry generation, ideally extending one of those frameworks to generate text in a way that is novelly sensitive to some socio-cultural factor such as politeness, blame, sentiment, etc. This will represent a novel technical advancement of interest to the natural language processing community as a whole, as well as the computational poetry generation community specifically. In the second phase of the project, the student will leverage their techniques to create a set of found poems, culminating in a unique performance at the 2020 High Grade literary journal release, in which selected poems written by the student will be co-read by the student and a machine or robot. Depending on project success, a paper on this project will be submitted to the International Conference on Computational Creativity, and a paper, video, and/or demo will be submitted to the International Conference on Human-Robot Interaction.

Elucidate the interdisciplinary nature of the project

This is a highly interdisciplinary project at the intersection of computer science and the humanities. As described above, the project will involve both technical and artistic components.

More Information

http://www.aclweb.org/anthology/W17-3502

http://www.johnbevis.com/pdf/Found_Poetry.pdf

Grand Engineering Challenge: Reverse-engineer the brain

Student Preparation


Qualifications

The student should taken Probability and Statistics and have experience with poetry writing. The student would benefit from having taken or co-registered in Machine Learning, Artificial Intelligence, or Human-Robot Interaction.

Time Commitment

20-40 hours/month

Skills/Techniques Gained

The student will gain experience with new CS skills in Machine Learning and Natural Language Processing, and new Poetry-writing skills in Found Poetry creation.

Mentoring Plan

The student will meet on a regular (weekly or bi-weekly) basis with Tom Williams and/or Toni Lefton, and will be invited to MIRRORLab meetings to engage with students working on other aspects of language generation.