metadata

language: he
tags:
  - hebrew
  - ocr
  - pos-tagging
  - historical-newspapers
  - ha-tsfira
license: mit
size_categories:
  - 10K<n<100K
task_categories:
  - text-classification
language_creators:
  - found
annotations_creators:
  - machine-generated
source_datasets:
  - original
pretty_name: Ha-Tsfira OCR and POS-Tagged Dataset

Ha-Tsfira OCR and POS-Tagged Dataset

Dataset Description

Repository: https://huggingface.co/datasets/mbole/hebrew-tzfira-dataset
Code and Tools: https://github.com/m-bole/hebrew_newspaper_processor
Point of Contact: Maria Bolek ([email protected])

Dataset Summary

This dataset contains OCR-processed and POS-tagged text from Ha-Tsfira (הצפירה), a Hebrew-language newspaper published in Poland from 1862 and then from 1874 to 1931. The dataset includes 50 newspaper issues that have been digitized, cleaned, and linguistically annotated.

Languages

Hebrew (he)

Dataset Structure

DatasetDict({
    train: Dataset({
        features: ['id', 'ocr_text', 'cleaned_text', 'pos_tagged'],
        num_rows: 50
    })
})

Data Fields:

id: String identifier (newspaper issue date)
ocr_text: Raw OCR output
cleaned_text: Normalized and cleaned text
pos_tagged: Text with part-of-speech tags

Dataset Size:

Total size: 15.2 MB
Download size: 6.7 MB
Number of examples: 50

Data Collection and Preprocessing

Source: National Library of Israel digital archives (https://www.nli.org.il/he/newspapers/hzf)
OCR: Tesseract OCR with Hebrew language support
POS Tagging: Stanza Hebrew model
Processing Steps:

PDF page splitting into columns
OCR using Tesseract
Text cleaning and normalization
POS tagging

Code and Tools

The code used to create this dataset is available on GitHub: Hebrew Newspaper OCR and POS Tagging Processor

Features include:

Automatic page splitting
Hebrew OCR processing
Text cleaning
POS tagging
Batch processing support

See the GitHub repository for installation and usage instructions.

Citation

bibtexCopy@misc{bolek2025hatzfira, title={OCR and Corpus Analysis of Ha-Tsfira Hebrew Newspaper}, author={Maria Bolek}, year={2025}, publisher={University of Warsaw}, howpublished={https://huggingface.co/datasets/mbole/hebrew-tzfira-dataset} }

Licensing Information

MIT License