Heads up!
The procedure for determining the similarity of documents used in CopySpider can not be interpreted as plagiarism measuring.
On the definition of plagiarism, we indicate the study of plagiarism.org content.
The 3% limit embedded in CopySpider aims to indicate possible existence of undue copies (plagiarism) between documents. This threshold is supported by international research on anti plagiarism and document similarity techniques.
It is important to note that the 3% limit is not directly related to the content comparison, word and paragraphs, from input files. For example, if you find a score of 10% between input file A and a internet found document B, it does not mean that 10% of document B has been found in the content of document A.
In simple terms, the 3% limit is related to the comparison and identification of rare chunks between the documents A and B. The rare chunks are sequences of words that occur a few times, mainly just once, in each document.
When CopySpider finds 3% or more of these rare chunks, international research conclude that there is much chance of those documents have mutual copies, that can be understood as plagiarism.
The conclusion about the error or lack associated with plagiarism has the user's (assessor) total responsibility when analysing of the results determined by CopySpider.
For more details we share some references that support the 3% threshold:
TEIXEIRA, C. M.; CICOGNA, M. A.; MORAIS, M. R. Software para detecção de textos com plágio baseado em busca pela internet. Anuário da Produção de Iniciação Científica Discente, [S.l.], v. 14, n. 24, p. 221-241, 2014. Disponível em: <http://sare.anhanguera.com/index.php/anuic/article/view/7564/1703>. Acesso em: 20 maio 2014.
LYON, C. M.; MALCOLM, J. A.; DICKERSON, R. G. Detecting short passages of similar text in large document collections. 2001. Disponível em: <https://uhra.herts.ac.uk/dspace/bitstream/2299/1695/1/901890.pdf>. Acesso em 20 maio 2014.