Reddit Cross-Topic Authorship Verification Corpus

Published: 18 July 2016| Version 1 | DOI: 10.17632/hppkn5kbg8.1
Oren Halvani


The "Reddit Cross-Topic Authorship Verification Corpus" consists of comments written between 2010 to 2016 from 1,000 reddit users. Each problem includes 1 unknown and 4 known documents (~ 7 KByte per document), where each document represents an aggregation of reviews coined from the same so-called subreddit. More precisely, all documents within a problem are disjunct regarding the subreddits in order to enable a cross-topic corpus. All subreddits cover exactly 1,388 different topics such as books, news, gaming, music, movies, etc. The corpus follows excatly the same format as the well-known PAN Authorship Identification corpora (



Linguistics, Machine Learning, Forensic Analysis