---
license: mit
task_categories:
- text-classification
- token-classification
- fill-mask
- text-generation
language:
- he
pretty_name: Sofer Mahir
---
# sofer_mahir
HTR project on big manuscripts of Rabbinic treatises from the Tannaitic period
(c) 2021 Daniel Stökl Ben Ezra (EPHE, PSL) and Hayim Lapin (University of Maryland)

Licence https://creativecommons.org/licenses/by-nc-sa/4.0/

Models are for use with kraken and/or eScriptorium (see https://escripta.hypotheses.org for further information).

If you are using the models, please quote: Stökl Ben Ezra, D., Brown-DeVost, B., Jablonski, P., Kiessling, B., Lolli, E., Lapin, H. “BiblIA – a General Model for Medieval Hebrew Manuscripts and an Open Annotated Dataset” HIP@ICDAR 2021.

Please upload your data and trained models according to the SA licence to enable improving the models for everybody. 
Transcription models here: 
1. Generalized Medieval Hebrew: (https://zenodo.org/record/5468286)
2. Ashkenazi_01 (https://zenodo.org/record/5468478)
3. Italian_01 (https://zenodo.org/record/5468573)
4. Sephardi_01 (https://zenodo.org/record/5468665)

Segmentation models will remain here until kraken's model repository on zenodo permits segmentation model upload.


BiblIAlong02_se3_2_tl.mlmodel permits to segment the main text regions and lines of manuscripts and strives to ignore marginal comments, commentary etc.
SoferMahirCleanFL06Eb_83_tl.mlmodel segments into "Main", "Margin", "Paratext" regions and distinguishes interlinear lines ("Correction")