2026 Conference Programs

Session

W02: Hallucinating the Hand: Developing an Evaluative Methodology for the Accuracy of Machine-Generated Handwritten Text Transcription

Time:

Wednesday, 14/Oct/2026:

12:00pm - 12:50pm

Location: Breakout 2

Presentations

Hallucinating the Hand: Developing an Evaluative Methodology for the Accuracy of Machine-Generated Handwritten Text Transcription

Kiley Jolicoeur¹, Mehr Athar^1,2

¹Syracuse University, United States of America; ²Indiana University, Bloomington, United States of America

Brief Abstract

The Department of Digital Stewardship at Syracuse University Libraries developed a computational methodology based in character error rate (CER) analysis to evaluate tabular handwritten text transcriptions produced by generative AI tools. These tools were tested on historical cemetery registers to explore opportunities for rapid, scalable transcription workflows.