Plagiarism detection supported by web usage mining for Turkish textual documents

The aim of this thesis is to compare the similarities among textual documents in Turkish, specifically student assignments from Çukurova University Department of Computer Engineering Moodle web site, and to detect plagiarism with a string matching algorithm. Since existing systems are developed for English and there are not many systems for Turkish, we designed this study for the Turkish language. We provide the detection of similarities when a change was done in the annexes of the words, the synonyms of words were used or phrase, sentence or paragraph were transposed. We also aim to improve the results of the similarity detection by analyzing the behavior of the students on Moodle using web usage mining methods. This study is focused on the solutions for detecting plagiarism which has increased due to the Internet usage in the academic environment. In the proposed method student assignments were compared with each other using the Greedy String Tiling Algorithm and Cosine Similarity methods. We observed that using the plagiarism detection tool reduces time and effort needed for the assessment of the textual assignments, prevents our students from plagiarizing, and increases their success in their course. Our proposed plagiarism detection tool was compared with the existing tools and methods, and its success was evaluated over sample texts.