An Experiment in Plagiarism Detection in Academic Research Articles Using Attributional Techniques

Abstract

There are certain overlapping aspects that brings plagiarism detection and authorship attribution together in one basket. Suspected cases of plagiarism can be interpreted as special cases of disputed or misattributed  authorship. Being so, the same techniques used to resolve doubtful cases of attribution can be used to investigate any potential existence of plagiarism. Principal components analysis and cluster analysis (henceforth PCA and CA) are among the popular statistical techniques used to proceed with various  attributional scenarios. These two techniques are used throughout this paper to explore the patterns of function words displayed in seventeen samples of one specific genre (academic research articles).

A survey is conducted over various cases of academic attribution: academic writing in English as a First Language and as a Second Language, and even cases of research articles with mixed authorship. Function words have been targeted in this paper as possible indicators of the author's identity. Accordingly a set of English function words is tested using WordSmith Tools (version 5.0). It turned out that the multivariate techniques (represented by PCA and CA) are most likely robust for addressing the type of issues raised about plagiarism and authorship attribution. Besides, it appeared that the statistical patterns of function words usage are rather relevant markers to deal with various scenarios of potential cases of plagiarism. This could explain the three different clusters plotted in the data environment for Halliday's samples, the Phillippine's samples together with his collaboratively authored ones, and the Iraqi's suspected samples that represented a highly potential case of plagiarism.

Keywords: Authorship Attribution, Plagiarism Detection, Authorship Verification