NHLOCAL/judaic-texts-corpus

Judaic Texts Corpus

Dataset Summary

This dataset is a comprehensive corpus of classic Jewish rabbinic literature, providing a vast collection of foundational texts in a clean, machine-readable format. The corpus spans a wide range of genres and historical periods, from the Tanakh and Talmud to Halakha (Jewish Law), Kabbalah, Chassidut, and modern rabbinic responsa.

The content is sourced from the otzaria-library project, curated by Sivan22, and was processed and structured for NLP applications from the NHLOCAL/otzaria-library fork.

The primary goal of this dataset is to provide a structured and accessible resource for training and evaluating language models, as well as for digital humanities research on Hebrew and Aramaic texts.

Topics Covered

The dataset is organized hierarchically and includes texts from the following major categories:

• תנ"ך (Tanakh)
• משנה (Mishnah)
• תוספתא (Tosefta)
• תלמוד בבלי (Babylonian Talmud)
• תלמוד ירושלמי (Jerusalem Talmud)
• מדרש (Midrash)
• שו"ת (Responsa)
• הלכה (Halakha - Jewish Law)
• ספרי מוסר (Musar - Ethical Literature)
• קבלה (Kabbalah)
• חסידות (Chassidut)
• מחשבת ישראל (Jewish Philosophy)
• סדר התפילה (Liturgy)
• ספרות עזר (Reference Works)

Dataset Structure

Data Fields

Each entry in the dataset corresponds to a single text file and is defined by the following features:

• book_name (string): The name of the book, derived from the source filename (e.g., מסכת ברכות).
• content (string): The full text content of the book. The text is in a pseudo-HTML format.
• category (string): The hierarchical category path from the original repository, preserving the structural organization (e.g., תלמוד בבלי/סדר זרעים).
• source_path (string): The full relative path to the original source file in the repository (e.g., אוצריא/תלמוד בבלי/סדר זרעים/מסכת ברכות.txt).

Data Instance

A typical data instance looks like this:

{
  "book_name": "מסכת ברכות",
  "content": "<h1>שם הספר: תלמוד בבלי</h1>\n<h2>שם המסכת: מסכת ברכות</h2>\n<h3>דף ב עמוד א</h3>\n<h4>גמרא</h4>\nמאימתי קורין את שמע בערבית? משעה שהכהנים נכנסים לאכול בתרומתן. עד סוף האשמורה הראשונה, דברי רבי אליעזר...",
  "category": "תלמוד בבלי/סדר זרעים",
  "source_path": "אוצריא/תלמוד בבלי/סדר זרעים/מסכת ברכות.txt"
}

Data Splits and Size

The dataset consists of a single split, train. No predefined validation or test splits are provided.

• Configuration: default
• Split: train
• Number of examples: 7,035+
• Download size: 1.72 GB

How to Use

You can load the dataset using the datasets library:

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("NHLOCAL/judaic-texts-corpus")

# Access the data
print(dataset['train'][0])

Dataset Creation

Curation Rationale

The corpus was created to consolidate a vast and scattered collection of essential Jewish texts into a single, unified, and programmatically accessible format. This facilitates large-scale analysis and the development of natural language processing models specialized in historical Hebrew and related dialects.

Source Data

The data was collected and digitized by the creators of the original otzaria-library repository. The text files were organized in a hierarchical folder structure, which has been preserved in the category and source_path fields. The data was then processed using an automated script to parse the file structure and content into the current tabular format.

Considerations for Using the Data

Limitations

The source texts are formatted using a simple, pseudo-HTML tagging system. While largely consistent, there may be formatting artifacts or parsing inconsistencies. Furthermore, as with any large digitized corpus, there may be occasional OCR errors or typographical mistakes from the digitization process.

Additional Information

Licensing Information

The dataset is released under the Apache License 2.0.

Please note that while the dataset compilation is under the Apache 2.0 license, the individual texts within the corpus may be subject to their own original licenses from their respective sources.

Citation Information

If you use this dataset in your research, please cite it as follows:

@dataset{judaic_texts_corpus,
  author = {NHLOCAL},
  title = {Judaic Texts Corpus},
  year = {2025},
  publisher = {Hugging Face},
  version = {1.1.0},
  url = {https://huggingface.co/datasets/NHLOCAL/judaic-texts-corpus}
}

Contributions

A special thank you to Sivan22 (Sivan22 on GitHub) for the immense work of collecting, digitizing, and organizing the source material in the original otzaria-library repository. This dataset would not be possible without their foundational effort.

The otzaria-library project, in turn, incorporates texts from several other invaluable digitization initiatives. We extend our gratitude to the following projects for making these texts publicly available:

• Sefaria (sefaria.org)
• Dicta (Dicta Library on GitHub)
• Orayta Books (Orayta-Books on GitHub)
• OnYourWay (ובלכתך בדרך) (mobile.tora.ws)
• Torat Emet (תורת אמת) (toratemetfreeware.com)