Efficiency of national and international software in detection of similarity and plagiarism in manuscripts

This study aims to identify the efficiency of fourteen software for detecting similarities in a text with fragments on financial education, found on internet websites, proceedings of academic events and open and restricted access scientific journals. The fragments were used to create a cohesive text, written in Portuguese, containing literal textual reproductions, paraphrases, with excerpts translated from the English language, manipulated with hidden character insertion and replacement of words by synonyms. The similarity reports generated by the software were analyzed according to four criteria: 1 identification of correctly cited fragments; 2 identification of plagiarized fragments; 3 identification of texts manipulated to deceive the software; 4 identification of original sources; which were scored on a scale of 0 to 4 points. The software Tumitin, StrikePlagiarism, PlagScan and Plagium had high performance and CopySpider and Plagium (add-on to Google) were the most inefficient. It was found that the software operates with variable efficiency, which reinforces the conviction that although they are useful tools for the identification of plagiarism, they contribute in a limited way to aspects such as the recognition of literal textual reproductions that do not always correspond to the original source. The original sources of rewritten, manipulated and translated texts were not found by any software. The study contributes to the improvement of the user's ability to choose, use and analyze the similarity reports generated by software, whose efficiency can be greater in the case of using more than one software.