Epstein Email Threads: 5,082 Parsed Communications

5,082 email threads extracted from documents released by the U.S. House Oversight Committee. The dataset has been processed with large language models to structure information including senders, recipients, timestamps, subjects, and message bodies.

Use Cases

• Perform network analysis on senders and recipients to map communication patterns.
• Train a text classification model on message bodies to categorize email content.
• Conduct temporal analysis using timestamps to reconstruct event sequences.
• Build a question-answering system using structured subjects and message bodies.

Strengths

• Contains 5,082 parsed email threads, providing a substantial text corpus.
• Data is structured and machine-readable, with OCR errors corrected and footers removed.
• Emails originate from an official U.S. House Oversight Committee document release.

Limitations

• The original OCR process may have introduced residual errors not fully corrected.
• Dataset scope is limited to a specific document set, not the complete email history.
• Potential class imbalance or bias inherent in the original document selection.

Provenance

Source

Documents released by the U.S. House Oversight Committee.

Collection Method

Extracted from OCR'd documents and processed using large language models for structuring.

Time Range

null

Freshness

null

Geography

Region is tagged as 'us'.