Bilingual dataset improves detection of cross-language plagiarism and AI text reuse
PlagiaCorpus: A Bilingual, Adversarial, and Cross-Lingual Dataset for Forensic Plagiarism Detection
Computation and LanguageArtificial IntelligenceComputers and SocietyDigital LibrariesInformation Retrieval
Summary
Finding copied or reused text is hard, especially when the text is changed a little or moved between languages. To help with this, the authors created PlagiaCorpus, a large collection of Arabic and English text pairs that includes examples of exact copying, tricky paraphrasing, machine-generated paraphrasing, and translated reuse. This dataset lets people build and test tools that catch plagiarism and reused text, even when AI changes or language differences are involved. It keeps clear training and testing groups so that tools can be fairly evaluated.
What this means in practice
- •For plagiarism detection teams: Develop and test tools that recognize copied and reworded content across Arabic and English, including AI-generated paraphrases and translations.
- •For content moderation teams: Improve automated detection of reused and translated text to maintain content originality in multilingual platforms.
Authors
Adel Ammar, Serry Sibaee, Wadii Boulila
Abstract
PlagiaCorpus is a bilingual Arabic-English dataset for benchmarking plagiarism and text-reuse detection, including reuse produced by large language models (LLMs) and reuse across languages. It contains 198,643 text records: a source corpus of 34,885 passages drawn from English and Arabic Wikipedia and two openly licensed book collections, and 163,758 queries (147,381 for training and 16,377 for testing). The queries cover five reuse classes of increasing difficulty, namely verbatim copying, character-level obfuscation, same-language paraphrasing by two different LLMs, and English--Arabic translation with structural paraphrasing, together with a Safe class of non-reused passages. Every positive query is linked to its source passage, and the train/test split is grouped by source passage so that no passage contributes queries to both splits. The release was curated through language identification, removal of LLM generation artefacts, and exact and cross-lingual de-duplication. Technical validation reports per-class length statistics, automatic checks of the train/test split, and inter-annotator agreement on a manually rated sample. The dataset is distributed in JSON and Parquet formats under a CC BY-SA 4.0 license.