The Text Score Dataset 1.0, Jennifer walshe

Composer Jennifer Walshe has been at the forefront of experiments with AI, art and music for over a decade. Beginning in 2017, Walshe embarked on a somewhat quixotic enterprise, the creation of a massive - one might say definitive, or at least definitively chaotic - dataset of text scores. Her goal was to train a machine learning system to generate new text scores so people could perform them. Walshe’s dataset now comprises over 3,000 text scores, running to almost half a million words, ranging from Fluxus event scores to recent works. Outputs from the project have been performed all over the world. 

The Text Score Dataset 1.0, first published online in 2021, and brought out in print for the first time in 2026 by RBO/Shift, features essays by Walshe and chief “data wrangler” Ragnar Olafsson, as well outputs from the project. Walshe reflects on how the act of dataset curation is never a neutral one, rife with omissions, compromises and endless artistic, technical legal and ethical decisions, making the curation of a dataset a singular and creative act. 

Selfpublished
Softback
140 x 215 mm
96 pages
SBN 9781915743121