Tobacco Lawsuit Documents with 7 Million Scanned Pages

Approximately 7 million scanned legal documents from the 1990s state lawsuits against the tobacco industry, released as part of the Master Settlement Agreement. The collection includes original TIFF page scans and text from 1990s-era OCR, featuring challenging elements like handwriting and stains. It was compiled by the Illinois Institute of Technology for research on complex document processing.

Use Cases

• Train document image analysis models on TIFF scans to detect and segment challenging features like handwriting and stains.
• Benchmark modern Optical Character Recognition systems against the provided 1990s-era OCR text to measure accuracy improvements.
• Develop information retrieval systems for legal archives using the document collection from the TREC Legal Track.
• Analyze document structure and layout in a large-scale corpus of real-world, 'messy' legal documents.

Strengths

• Large scale of approximately 7 million documents provides substantial data for training.
• Includes both original TIFF page scans and derived OCR text, offering multiple data modalities.
• Documents contain real-world complexities like handwriting and stains, useful for robust model testing.

Limitations

• OCR text was generated with 1990s technology, likely containing significant noise and errors.
• The dataset's content is specific to a single legal case (tobacco lawsuits), limiting generalizability.
• Document scans are in TIFF format, which may require conversion for some modern vision pipelines.

Provenance

Source

Illinois Institute of Technology (IIT), from documents made public via the Master Settlement Agreement and archived at UCSF.

Collection Method

Documents were gathered from the public archive resulting from the tobacco industry lawsuit settlement.

Time Range

1990s (period of the lawsuits and settlement).

Freshness

Last updated in July 2022, but the underlying documents are from the 1990s.

Geography

United States (from lawsuits filed by U.S. states).