HYBRID ARABIC TEXT PLAGIARISM DETECTION SYSTEM رسالة ماجستير

Abstract

This research presents a Hybrid Arabic Plagiarism Detection System (HYPLAG). HYPLAG is an Arabic text-based plagiarism detection approach that combines corpus-based and knowledge-based methods by utilizing Arabic semantic resource. The main aim of this research is to find out the effect of the combining process on the performance of detection process on Arabic plagiarized text cases. A preliminary study on undergraduate students was conducted to understand their behaviors or patterns in plagiarism. The results of the study show that students apply different methods to plagiarized sentences, also it shows changes in sentence’s components (verbs, nouns, and adjectives). Based on these results, HYPLAG was developed taking into account other patterns of plagiarism. Given a suspicious document and a collection of documents, HYPLAG compares the input document against the document collection in an efficient manner. It utilizes the search engine structure in the retrieving method, where the most relevant source sentence is retrieved. To ensure the validity of the input text, a set of preliminary methods are applied. One of most important methods is the stemming, where terms are replaced with their original roots. To choose an accurate stemmer from a set of Arabic current proposed stemmers, a comparative approach based on a lexicon resource is proposed. Evaluating HYPLAG against several other approaches demonstrates its higher performance with less computational time using the same dataset “Extrinsic Arabic Plagiarism Detection Dataset, ExAraPlagDet-2015”. HYPLAG achieves a precision and recall values of 92% and 87%, respectively.