Historical Newspaper Images With Extracted People Data

Newspaper Navigator People is a dataset derived from the Library of Congress's Chronicling America newspaper archive. It contains images and text extracted from historical US newspapers, specifically focusing on visual content labeled as containing people. The dataset was created by davanstrien and last updated in March 2022.

Use Cases

• Train object detection models to identify images containing people within scanned newspaper pages.
• Perform named entity recognition on the extracted text captions or surrounding OCR text associated with the image crops.
• Analyze temporal trends in newspaper photography or illustration styles using the image data and associated metadata.
• Build multimodal models that link visual content of people with textual context from the original newspaper articles.

Strengths

• Sourced from the authoritative Library of Congress Chronicling America digital newspaper collection.
• Contains multimodal data (images and text) enabling joint analysis.
• Platform tags indicate a focus on Named Entity Recognition and Optical Character Recognition tasks.

Limitations

• Specific row count, column names, and dataset size are unknown, limiting assessment of scale.
• Data is derived from OCR, which may contain recognition errors affecting text quality.
• Last update was in 2022, so the dataset may not reflect the most recent state of the source archive.

Provenance

Source

Library of Congress Chronicling America historical newspaper archive.

Collection Method

Images and text extracted via Optical Character Recognition (OCR) and likely machine vision models from newspaper scans.

Time Range

null

Freshness

Last updated 2022-03-16.

Geography

United States (Region:us per platform tags).