A Novel Approach for Analysis of ‘Real World’ Data: A Data Mining Engine for Identification of Multi-author Student Document Submission

Abstract

In this article we describe a data mining engine which makes use of anew approach to plagiarism detection. The new approach which we havetaken identifies student submissions which have been produced by morethan one author and hence provides a starting point for investigation ofa student submission which may contain plagiarized material. Theapproach, which this engines uses, has great potential for use by thosemarking submissions from two types of student bodies. Namely large classsize with whose written styles they may not be familiar and studentsfollowing online courses who they may not ever meet. The approach whichwe have taken is new in that other approaches endeavor to match thesubmitted material with material existing elsewhere whereas our approachattempts to determine multiple author styles in the submission and henceprovide an indication that the submission contains information from morethan one source. The implications of the use of author styles foridentification of future suspect submissions, and for comparison withfuture submissions by the same student, are discussed.