Datasets:

NHLOCAL
/

project-ben-yehuda

license: cc0-1.0
language:
  - he
tags:
  - hebrew
  - text-corpus
  - public-domain
  - literature
  - historical-texts
  - project-ben-yehuda
pretty_name: Project Ben-Yehuda Corpus
dataset_info:
  download_size: null
  dataset_size: null
size_categories:
  - 10K<n<100K

Project Ben-Yehuda Corpus

Dataset Description

This dataset contains a comprehensive collection of public domain Hebrew texts, sourced from the digital library of Project Ben-Yehuda. The corpus includes thousands of works spanning various genres such as prose, poetry, essays, and translations, primarily from the Haskalah period to the early 20th century.

The data has been structured and cleaned to be readily usable for Natural Language Processing (NLP) tasks, including language modeling, text analysis, historical linguistics research, and more. Each entry in the dataset corresponds to a single literary work and includes both the full text and rich metadata from the project's catalog.

Dataset Structure

The dataset consists of a single split, train. Each record contains three main fields: text, source, and metadata.

Data Instances

A typical data instance looks like this:

{
  "text": "לְפֶתַח הַסִּפְרִיָּה הָעִירוֹנִית הַגְּדוֹלָה, מִבְּנֵי-הַבְּרִיחַ הָאַחֲרוֹנִים לְאוֹתָהּ הָעוֹנָה, אֲשֶׁר נִשְׁאַר בָּעִיר לִרְפוּאָה, הָיָה מִתְגַּנֵּב יוֹם-יוֹם בְּשָׁעָה קְבוּעָה וּבְדִיּוּק גָּמוּר אָדָם אֶחָד, עוֹבֵר-אֹרַח, שֶׁאִישׁ לֹא יְדָעוֹ. הוּא הָיָה בָּא תָּמִיד בִּשְׁעַת-הַצָּהֳרַיִם...",
  "source": "Project Ben-Yehuda",
  "metadata": {
    "pby_id": 18,
    "title": "פת לחם",
    "authors": "יוסף חיים ברנר",
    "translators": "",
    "author_uris": "https://wikidata.org/wiki/Q939732",
    "translator_uris": "",
    "original_language": "",
    "genre": "פרוזה",
    "source_edition": "ע\"פ “כתבים”, כרכים א‘-ב’, תל-אביב: הקיבוץ המאוחד, תשל“ח-תשמ”ה",
    "filepath_pby": "/p66/m18"
  }
}

Data Fields

text (string): The full, unprocessed text content of the work.
source (string): The origin of the data, consistently set to "Project Ben-Yehuda".
metadata (struct): A nested object containing detailed information about the work:
- pby_id (int64): The unique identifier for the work in the Project Ben-Yehuda catalog.
- title (string): The title of the work.
- authors (string): The name(s) of the author(s).
- translators (string): The name(s) of the translator(s), if applicable.
- author_uris (string): Wikidata URIs for the author(s).
- translator_uris (string): Wikidata URIs for the translator(s).
- original_language (string): The original language of the work, if it's a translation.
- genre (string): The literary genre (e.g., "פרוזה", "שירה", "מאמרים ומסות").
- source_edition (string): Information about the printed edition used as the source for the digital text.
- filepath_pby (string): The original file path in the Project Ben-Yehuda archive structure.

Dataset Creation

The dataset was created by processing the publicly available files from Project Ben-Yehuda. The process involved:

Parsing the pseudocatalogue.csv file provided by the project to extract metadata for each work.
Matching each entry in the catalog to its corresponding text file based on the provided file path.
Reading the content of each text file.
Combining the text and metadata into a structured JSONL format, sharded into multiple files to adhere to repository size limits.
Finally, an automated workflow converts the entire collection into a single, efficient Parquet file for use with the datasets library.

The scripts used for this process are available at https://github.com/NHLOCAL/The-Digital-Genizah.

Source Data and Attribution

This dataset is made possible by the incredible work of Project Ben-Yehuda and its many dedicated volunteers, who have worked for decades to digitize and make Hebrew literature accessible to all. We extend our deepest gratitude to them. For more information about the original project, please visit their website: benyehuda.org.

Disclaimer: This dataset is an independent initiative and is not officially affiliated with, or endorsed by, the official Project Ben-Yehuda. It was created solely for the purpose of making this valuable public domain resource more accessible for computational research and NLP applications in a standardized, machine-readable format. The text content is presented as-is from the source files.

Licensing Information

The texts sourced from Project Ben-Yehuda are in the public domain. In the spirit of open access and to maximize usability, this compiled dataset is released under the Creative Commons Zero v1.0 Universal (CC0 1.0) Public Domain Dedication. You can copy, modify, distribute, and perform the work, even for commercial purposes, all without asking permission.

Citation Information

If you use this dataset in your research, please consider citing the original Project Ben-Yehuda as the source of the texts. You can also cite this dataset to ensure reproducibility:

@dataset{project_ben_yehuda_corpus_nhlocal,
  author       = {nhlocal and Project Ben-Yehuda},
  title        = {Project Ben-Yehuda Corpus},
  year         = {2025},
  publisher    = {Hugging Face},
  dataset      = {nhlocal/project-ben-yehuda},
  url          = {https://huggingface.co/datasets/nhlocal/project-ben-yehuda}
}