Scholars from Vilnius University in Lithuania have launched a breakthrough AI tool capable of reading handwritten Yiddish. The Vilne-Yiddish model, developed at the university’s joint Digital Humanities Laboratory, was created by Sergii Gurbych, of the Center for the Study of East European Jewry. It seeks to transform access to Jewish archival materials, drawing on autobiographies submitted to YIVO in the 1930s and recently rediscovered in Lithuanian archives.

Gurbych said the different handwriting styles “differ by period, region and even by social background,” and that manually transcribing texts is “time-consuming and labor-intensive.”

The tool reaches roughly 95% accuracy. “Post-recognition manual correction remains necessary due to inevitable errors,” he added. The model is open-access, Gurbych said, emphasizing that researchers can adapt the model to new handwriting styles using only a dozen pages of text.