Mining Naturally-occurring Corrections and Paraphrases from Wikipedia's Revision History

Naturally-occurring instances of linguistic phenomena are important both for\ntraining and for evaluating automatic processes on text. When available in\nlarge quantities, they also prove interesting material for linguistic studies.\nIn this article, we present a new resource built from Wikipedia's revision\nhistory, called WiCoPaCo (Wikipedia Correction and Paraphrase Corpus), which\ncontains numerous editings by human contributors, including various corrections\nand rewritings. We discuss the main motivations for building such a resource,\ndescribe how it was built and present initial applications on French.\n

Paper

References (12)

1113 Wikipedia is available in as many as 269 languages as of March, 2010, and our techniques can be applied on all languages with mostly clear word segmentation13 Wikipedia is available in as many as 269 languages as of March, 2010, and our techniques can be applied on all languages with mostly clear word segmentation
12A list of words built by applying the most frequent edition scripts to the word to correct (denoted “patterns”)

Similar papers

© 2026 NYSGPT2525 LLC