Usenet Corpus 1980–2013: Over 300 Billion Tokens of Human Discourse

OwnedByDanes's Usenet Corpus 1980–2013 is a deduplicated and sanitized archive of Usenet posts. It contains over 300 billion tokens of human discourse spanning three decades of internet history, with the full corpus comprising 103.1 billion tokens across 408 million posts. A sample of 65,000 rows is hosted on the platform.

Use Cases

• Train large language models based on the extensive, multi-decade text corpus.
• Analyze linguistic and cultural trends over time based on three decades of human discourse.
• Study the evolution of online communities and topics based on Usenet discussion threads.
• Benchmark text generation or classification models based on historical internet text data.

Strengths

• Over 300 billion tokens of text data provide substantial volume for training.
• Covers a three-decade time range from 1980 to 2013 for longitudinal analysis.
• The full corpus is deduplicated and sanitized, which likely improves data quality.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The sample size of 65,000 rows is small relative to the full 408-million-post corpus, limiting initial inspection.
• Data may reflect temporal and platform bias inherent to the Usenet source.

Provenance

Source

OwnedByDanes

Collection Method

Likely gathered from public Usenet archives.

Time Range

1980–2013

Freshness

Last updated 2026-04-28 18:20:01; freshness should be verified.