Let's train computers to read ancient Hebrew manuscripts!
Tikkoun Sofrim is a joint French Israeli project aimed to make Medieval Hebrew manuscripts openly and freely available as texts. The project is combining automatic Handwritten Text-Recognition (HTR) and Crowdsourcing. The project is co-managed by the Department of Information Systems and the Department of Jewish Studies.
In a first stage we analyse the manuscript layout and train Kraken, a deep learning engine for automatic reading. Kraken is reading quite well, with an error rate of less than 10% and often even less than 3% on the letter level.
But this is not yet good enough.
In order to further improve Kraken’s automatic reading and provide efficient editions of the texts, we need the human eye.
The tool in this website is aimed at achieving this goal. Here you can check the automatic reading and correct mistakes.
In the next stage your corrections will be used for improving automatic reading as well as creating digital critical editions and enabling textual search of manuscripts in library viewers.
Go to Tikkoun Sofrim and start correcting!
____________________
Tikkoun Sofrim is a joined French-Israeli project, developed by the EPHE, PSL, in Paris, the eLijah-Lab at the University of Haifa and the National Library, Israel. The project is supported by the Maimonides grant funded by the French Ministry of Higher Education and Research, the French Ministry of Foreign Affairs and the Israeli Ministry of Science.
For questions, suggestions or any other remark, feel free to contact us