FanChuan: Multilingual Graph-Structured Parody Detection Benchmark

FanChuan is a multilingual, graph-structured benchmark containing between 10,000 and 100,000 records for parody detection and analysis on social media, published by Ziyi510 in 2025. The data covers six distinct subsets including Reddit-Trump, Tiktok-Trump, and CampusLife in both English and Chinese.

Use Cases

• Binary classification of parody vs. non-parody content in the Reddit-Trump subset
• Graph neural network modeling using the graph-structured relationships in the Tiktok-Trump data
• Cross-lingual parody analysis between English and Chinese subsets

Strengths

• Multilingual support for English and Chinese text
• Graph-structured format for relational parody analysis
• 10,000 to 100,000 records across six thematic subsets

Limitations

• Thematic bias towards specific topics like Trump or campus life
• Potential label noise common in social media parody benchmarks

Provenance

Source

Ziyi510, 'FanChuan: A Multilingual and Graph-Structured Benchmark For Parody Detection and Analysis' (arXiv:2502.16503)

Collection Method

Scraped from social media platforms including Reddit and TikTok

Freshness

Last updated March 2026 according to metadata.

Geography

Multilingual (English and Chinese)