PLAGIARISM A LA MODE - A COMPARISON OF AUTOMATED SYSTEMS FOR DETECTING SUSPECTED PLAGIARISM

Early automated systems for detecting plagiarism in student programs e mployed attribute counting techniques in their comparisons of program texts, while more recent systems use encoded structural information. W hales claims that the latter are more effective in their detection of plagiarisms than systems based on attribute counting, To explore the v alidity of these claims, a comparison is presented of five systems, tw o based on attribute counting and three using metrics based on structu re, The major result of this study is that the systems based on struct ural information consistently equal or better the performance of syste ms based on attribute counting metrics. A second conclusion is that of the structure metric systems, one using approximate tokenization of i nput texts (YAP) is as effective as a system that undertakes a complet e parse (Plague), Approximate tokenization offers a considerable reduc tion in the costs of porting to new languages, A distinction is also m ade between forms of plagiarism common among novice programmers and th ose employed by more experienced programmers.