Library of Congress Public Domain English Book Collection

Over 140,000 English books from the Library of Congress, comprising approximately 8 billion words, are included in this public domain collection. The dataset was compiled by Sebastian Majstorovic via the LoC JSON API, focusing on the Selected Digitized Books collection. It was last updated on the Hugging Face platform in March 2024.

Use Cases

• Train large language models on the 8 billion words of OCR text for generative tasks.
• Analyze historical language use and vocabulary trends across the full text of 140,000 books.
• Perform named entity recognition or topic modeling on the book contents to study cultural themes.
• Benchmark OCR post-processing algorithms using the raw digitized text from the collection.

Strengths

• Large scale with over 140,000 individual books.
• Substantial text volume of approximately 8 billion words.
• Public domain status ensures broad usability without copyright restrictions.

Limitations

• Specific row count, column details, and file formats are unknown.
• The dataset description indicates reliance on OCR, which likely introduces character recognition errors.
• Geographic and temporal coverage of the books is unspecified, limiting contextual analysis.

Provenance

Source

Library of Congress Selected Digitized Books collection.

Collection Method

Curated via the LoC JSON API, filtered for English books.

Time Range

null

Freshness

null

Geography

null