biglam/loc_beyond_words

Dataset Card for Beyond Words

Dataset Summary

The Beyond Words dataset is a crowdsourced collection of bounding box annotations on World War I-era historical newspaper pages from the Library of Congress’s Chronicling America collection. Volunteers marked seven types of visual content — photographs, illustrations, maps, comics, editorial cartoons, headlines, and advertisements — enabling the training of the visual content recognition model behind the Newspaper Navigator project. It serves as a foundational dataset for large-scale document layout analysis in historical archives.

Supported Tasks and Leaderboards

• Object detection
• Visual content classification
• Document layout analysis

Languages

• English (used in transcribed captions and OCR content)

Dataset Structure

Data Instances

Each instance is an image of a historic newspaper page, annotated with bounding boxes around regions containing visual content.

Data Fields

• image_id: Unique identifier for the image.
• image: Full page image from Chronicling America.
• width, height: Image dimensions.
• objects: List of annotations, each including:
  • bw_id: Unique Beyond Words annotation ID.
  • category_id: One of 7 class labels (Photograph, Illustration, etc.).
  • image_id: Reference to source image.
  • id: Object instance ID.
  • area: Area of the bounding box.
  • bbox: Bounding box coordinates (x, y, width, height).
  • iscrowd: Crowd label (for COCO format compatibility).

Data Splits

• Train: 2,846 examples
• Validation: 712 examples

Dataset Creation

Curation Rationale

To train models that can detect and classify visual content in historic newspapers at scale, the Library of Congress launched the Beyond Words crowdsourcing initiative in 2017. The data produced is designed to support machine learning workflows and historical content analysis.

Source Data

• Scanned WWI-era newspapers from Chronicling America
• Public domain metadata and images
• OCR from METS/ALTO XML files

Annotations

• Collected via the Beyond Words crowdsourcing platform
• Up to 6 volunteers per annotation; verified by consensus
• Categories include photograph, illustration, map, comic/cartoon, editorial cartoon
• Additional headline and advertisement annotations added by project team

Personal and Sensitive Information

None known. All pages are from historical public domain newspapers.

Considerations for Using the Data

Social Impact

• Democratizes access to historical newspaper content
• Supports digital humanities, education, and public history initiatives

Discussion of Biases

• Limited to WWI-era newspapers
• Class distribution skewed toward some categories (e.g. headlines, ads)
• Some annotations (headlines, ads) not crowd-verified

Limitations

• Visual content from pre-1875 newspapers may yield lower model performance
• Quality of annotations can vary due to the experimental nature of the crowdsourcing workflow

Additional Information

Dataset Curators

• Benjamin Charles Germain Lee
• Jaime Mears, Eileen Jakeway, Meghan Ferriter, Chris Adams, Nathan Yarasavage, Deborah Thomas, Kate Zwaard (Library of Congress)
• Daniel Weld (University of Washington)

Licensing

• License: CC0 1.0 Universal (Public Domain Dedication)

Citation

@inproceedings{10.1145/3340531.3412767,
  author = {Lee, Benjamin Charles Germain and Mears, Jaime and Jakeway, Eileen and Ferriter, Meghan and Adams, Chris and Yarasavage, Nathan and Thomas, Deborah and Zwaard, Kate and Weld, Daniel S.},
  title = {The Newspaper Navigator Dataset: Extracting Headlines and Visual Content from 16 Million Historic Newspaper Pages in Chronicling America},
  year = {2020},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  doi = {10.1145/3340531.3412767},
  url = {https://doi.org/10.1145/3340531.3412767}
}


### Contributions

Thanks to @davanstrien for adding this dataset to Hugging Face.