jmgronski · PyLaia · Published July 12, 2025
Letter Reader
Text Recognition
Description
Letter Reader focuses on early 20th century Yiddish, by authors with various levels of education and exposure to Yiddish. It handles informal orthography, and A wide range of regional vocabulary. It has been trained on: Family letters, Letters to relief societies - Grajewo Relief Society, Joint Distribution Committee (JDC), Minutes from Lithuanian Jewish community groups, and hand-written plays.
Ideal for: Social historians, family researchers working with private papers, archives with handwritten correspondence.
Credits This work was made possible by L’Dor V’Dor AI Lab Yiddish team with the generous support of the American Jewish Joint Distribution Committee (JDC), LitvakSIG, YIVO, and numerous individual volunteers who contributed documents and transcriptions. This model used the Dybbuk for Yiddish Handwriting model as a base developed by Sinai Rusinek and her team
For more information, please visit: https://ldvdf.org
Try this model
Use this modelOpen in Transkribus
Low error rate9.35% CER
Character Error Rate (CER) measures the percentage of characters incorrectly recognised. Lower is better. This model scored 9.35% on its validation set. As a rule of thumb, a CER below 10% is considered good for most handwritten material. This is a larger model trained on diverse material, which generally makes it more robust across different handwriting styles. That said, larger training sets also make it harder to push the CER down further.
Measured on the model's own validation data. Results on your documents may differ depending on handwriting style, document condition, language, and how closely your material resembles the training data.
Words99,526
Lines12,240
Training Pages532
Model ID371705