The 18th World Congress of Jewish Studies

eScriptorium: Using AI for Analyzing and Transcribing Manuscripts and Fragments

eScriptorium is a platform for automatic transcription of documents, handwritten, printed or inscribed. It is currently the only such cutting edge platfor to be fully open source including its interface and its artificial intelligence parts, trainable automatic layout segmentation and trainable transcription. Its ergonomic interface allows rapid creation of training data. Models and data can be published open source as we have done in a first publication (HIP@ICDAR 2021).

Through interaction with additional external tools allow the alignment with existing transcriptions, up to the full-fledge production of digital critical TEI-editions and their visualisation with TEI-Publisher.

We will present some of the results on literary manuscripts )Biblical and Rabbinic) as well as Qumran Genizah fragments from three projects, BiblIA, Sofer Mahir, SQE and HTR4PGP.

Marina Rustow, Hayim Lapin, Bronson Brown-Devost, Pawel Jablonski are coauthors of the presentation