Research

The Spatial Logic of Early Dynastic Cuneiform Writing: A Preliminary Analysis of Texts from Pre-Sargonic Lagaš

Developing a quantitative methodology for studying how the spatial relationships of cuneiform signs changes during one of the most foundational shifts in early writing.

SumTablets: A Transliteration Dataset of Sumerian Tablets

with Richard Diehl Martinez and Dan Jurafsky

Released a dataset mapping Unicode cuneiform glyphs to transliterations (91,606 tablets / 6,970,407 glyphs). Set naïve + transformer baselines on transliteration task, reaching 97.55 chrF on held-out transliteration.

ACL Anthology·arXiv·Repo·Dataset

Babbage in Babylon: Pioneering Deep Learning Approaches to Sumerian Cuneiform

with Dan Jurafsky, Ian Morris, and Niek Veldhuis

Deep learning approaches to three cuneiform tasks: OCR, transliteration, and translation. Awarded the Firestone Medal for Excellence in Undergraduate Research.

Stanford Digital Repository