A Plagiarism Detection System for Arabic Documents

This paper proposes a new plagiarism detection system devoted to Arabic text documents. This system is based on modeling the relation between documents and their n-gram phrases. Part-of-Speech tagging is applied on the examined documents to support in resolving the morphological ambiguity during text normalization. Text indexing and stop-words removal are performed, employing a new morphological analysis based method. Heuristic pairwise phrase matching algorithm is used to build the documents TF-IDF model, considering substitution of words with their synonyms. The hidden associations of the unique n-gram phrases contained in the documents are investigated using the Latent Semantic Analysis. Then, the pairwise document similarity scores are derived from the Singular Value Decomposition computations. The performance of the proposed system was confirmed through experiments with various data sets, exhibiting promising capabilities in identifying literal and some types of intelligent plagiarism. Finally, the proposed system was compared to Plagiarism-Checker-X, and the proposed system outperformed Plagiarism-Checker-X, especially for intelligent plagiarism.