Jokes: A Corpus of 700,161 Unique Text Jokes

Hudsongouge provides a collection of 700,161 unique jokes, deduplicated from a personal archive. The dataset was uploaded to Hugging Face and last updated on July 9, 2026. The author notes the jokes were found on a personal drive and their original source is unknown.

Use Cases

• Train language models for joke generation based on the large text corpus.
• Fine-tune classifiers to detect humor or inappropriate content based on the described joke text.
• Analyze linguistic patterns and structures in humorous text based on the provided examples.
• Benchmark text generation models on creative writing tasks using the joke collection.

Strengths

• Contains 700,161 unique jokes, indicating a large-scale text collection.
• Exact-text deduplication was performed on the raw entries, suggesting an effort to remove duplicates.

Limitations

• The author warns some jokes may not be good or could be inappropriate, indicating uncurated and potentially low-quality content.
• The original source and collection methodology are unknown, limiting provenance assessment.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

Personal archive of author 'hudsongouge'; original sources unknown.

Collection Method

Found as a 'jokes.json' file on a personal drive; exact-text deduplication applied.

Freshness

Last updated 2026-07-09 15:45:41; freshness should be verified.