Constructing an Academic Thai Plagiarism Corpus for Benchmarking Plagiarism Detection Systems

Plagiarism is a major problem in the academic world. It does not only undermine the credibility of educational institutions, but also interrupts the processes of creating knowledge in the academic community. To lessen this problem, many plagiarism detection systems have been developed to detect plagiarized texts in academic works. In this paper, we describe the design and process in creating an academic Thai plagiarism corpus. This corpus is necessary for training and testing plagiarism detection systems for Thai. In order to make this corpus a comprehensive representation of plagiarism, the data has been divided into various types based on the degree of the linguistic mechanisms used in plagiarism. Data compiled in our corpus comes through two main methods: manually created by participants and automatically generated by a program. After the corpus is created, its validity is verified by using three measurements: a measurement of similarity between suspicious texts at the character level, a measurement of similarity between suspicious texts at the word level, and a comparison of different types of data compiled in the corpus based on the similarity measured. The results of the analyses indicate that the corpus created by the proposed methods is effective in training and testing plagiarism detection systems.