Usando permutation based indexing na detecção de plágio

Abstract

Extrinsic text plagiarism detection is a document evaluation process, in which we analyze its content for possible plagiarism by comparing directly with potential source documents. The identification of extrinsic plagiarism can be divided into three stages, Heuristic Retrieval, Detailed Analysis and Postprocessing. This work will focus on the Heuristic Retrieval stage, and for that we will use the Permutation Based Indexing(PBI) approach, which was proposed as a new approach to the calculation of similarity between objects, having as a differential the reduction of the number of comparisons in the dataset, comparing the query only with the pivots objects, which are objects of the dataset itself chosen in the pivot selection step, considering to choose the objects that best represent the dataset as a whole. In addition to using the PBI technique, to add value to this work, this work will create variations on existing pruning techniques, based on a "pruning" of pivots, which removes pivots that do not have much influence on a given query.