Detection of Simple Plagiarism in Computer Science Papers

Proceedings of the 23rd International Conference on Computational Linguistics (Coling 2010), pages 421–429, Beijing, August 2010 Detection of Simple Plagiarism in Computer Science Papers Yaakov HaCohen-Kerner Department of Computer Science, Jerusalem College of Technology (Machon Lev) kerner@jct.ac.il Aharon Tayeb Department of Computer Science, Jerusalem College of Technology (Machon Lev) aharontayeb@gmail.com Natan Ben-Dror Department of Computer Science, Jerusalem College of Technology (Machon Lev) bd.natan@gmail.com Abstract Plagiarism is the use of the language and thoughts of another work and the repre- sentation of them as one's own original work. Various levels of plagiarism exist in many domains in general and in aca- demic papers in particular. Therefore, di- verse efforts are taken to automatically identify plagiarism. In this research, we developed software capable of simple plagiarism detection. We have built a corpus (C) containing 10,100 academic papers in computer science written in English and two test sets including pa- pers that were randomly chosen from C. A widespread variety of baseline me- thods has been developed to identify identical or similar papers. Several me- thods are novel. The experimental results and their analysis show interesting find- ings. Some of the novel methods are among the best predictive methods. 1 Introduction In light of the explosion in the number of availa- ble documents, fast and accurate searching for plagiarism is becoming more needed. Identifica- tion of identical and similar documents is becom- ing very important. Plagiarism is the use of the language and thoughts of another work and the representation of them as one's own original work (Wikipedia, 2010; Library and Information Services, 2010). Plagiarism can be committed by "recycling" oth- er's work as well as by one’s own work (self- plagiarism). Various levels of plagiarism exist in many domains in general and in academic papers in particular. In addition to the ethical problem, plagiarism in Academics can be illegal if copy- right of the previous publication has been trans- ferred to another entity. It is important to mention, that in many cases similar papers are different versions of the same work, e.g., a technical report, a poster paper, a conference paper, a journal paper and a Ph. D. dissertation. To avoid any kind of plagiarism, all sources which were used in the completion of a work/research must be mentioned (Library and Information Services, 2010). Over the last decade, various softwares have been built to automatically identify plagiarism (e.g., Collberg et al. (2005), Sorokina et al. (2006), and Keuskamp and Sliuzas (2007)). In this research, we developed such a system. This system is planned to deal with simple kinds of plagiarism, e.g., copying of sentences or part of sentences. We have built a corpus that con- tains academic papers in computer science writ- ten in English. Most of the papers are related to the domain research of Natural Language Processing (NLP) and are from the last ten years. The remainder of this paper is organized as follows: Section 2 gives a background regarding plagiarism. Section 3 overviews researches and systems dealing with detection …

Content truncated to 500 words