Many universities and academic journals now use plagiarism detection software, such as Turnitin®, to identify essays and research papers that have borrowed excessively from previously published work. However, in a blog post on BioMed Central, Ann Rogerson draws attention to a new type of paraphrasing software that may make this process less effective.
Paraphrasing tools, or ‘article spinners’, rephrase blocks of text so that they are no longer identical to the source material. In a study published in the ‘International Journal for Educational Integrity’, Rogerson and co-author Grace McCarthy applied two such paraphrasing tools – both freely available online – to an extract from one of their own previous publications. By comparing the resulting output to the original text, they showed that the programmes replace key words with synonyms, but typically avoid changing conjunctions or long strings of letters such as those found in URLs. Notably, Turnitin® failed to identify the output from either programme as plagiarism.
The paraphrased output contained a number of dubious word choices and grammatical errors that would probably ring alarm bells for journal editors and university lecturers. Nevertheless, machine learning will likely enable these programmes to become increasingly accurate over time. Rogerson and McCarthy therefore argue that it is important to have an open discussion about the implications – both ethical and practical – of using paraphrasing software. Researchers and students should receive training in how to describe others’ work in their own words with appropriate attribution. Finally, journal editors and universities should learn to recognise the red flags that may signal the use of paraphrasing tools.