Siamese GRU for Arabic-English Cross-language Plagiarism Detection

Plagiarism is a significant issue in academic and research fields, posing challenges to the integrity of scholarly work. With the increasing availability of multilingual resources, cross-language plagiarism has become a growing concern, where content is translated from one language to another to avoid detection. Detecting cross-language plagiarism is particularly challenging when it involves the Arabic language, given its rich morphology and complex syntax. This study proposes a novel approach for detecting cross-language plagiarism targeting the English-Arabic language pair. The methodology includes the pre-trained XLM-Roberta transformer combined with a Siamese Gated Recurrent Unit (SGRU) architecture, applied on the OPUS dataset, specifically utilizing data from Wikipedia. The experimental results demonstrate that this combination outperforms the state-of-the-art, achieving an accuracy of 95.52%, a precision of 94.93%, a recall of 96.13%, and an F1-score of 95.52%.