Handling Illusive Text in Document to Improve Accuracy of Plagiarism Detection Algorithm

Abstract

Plagiarism Detection is being one of the challenging tasks in the academic research world to ensure the integrity/authenticity of a document. Currently, many efficient algorithms are available to sufficiently detect plagiarism in a document. Pre-processing of a document typically remains a master key to achieving the maximum stable goal. Although all algorithms, before checking plagiarism, initially perform some sort of pre-processing on documents and convert the document into a particular format like by removing whitespaces and all special characters, etc. In this paper, we focus on two possible techniques, which can be used for plagiarism, which existing plagiarism detection algorithms are omitting. First is replacing the white spaces with a hidden character with white color (background color) between consecutive words so apparently, they seem to be distinct words, but the algorithm/computer will incorrectly consider them as a single word. So even a 100% copied statement would not be identified as plagiarized content. The second is hiding spam text behind images to falsely report the maximum number of words count in a document but as they are hidden so human eye can’t discover them and the algorithm will consider them as some words resulting in less percentile score of the plagiarized document. Our proposed (pre-processing) technique can efficiently handle these two critical problems which result in improved accuracy and authenticity of plagiarism checking algorithms. We have compared the performance of our algorithm considering these critical issues with other state-of-art algorithms (particularly with Turnitin) and our algorithm handles these issues efficiently.