The Crux of Reducing Similarity

Over the years, I've been contacted a dozen of times by clients who wanted me to reduce the similarity score of their academic research papers or theses. 

Although reducing similarity involves paraphrasing and falls within my scope as a line editor, I have to turn down certain projects. If the text in question is a highly scientific text I have to pass. This is because scientific texts (natural sciences, ICT, etc) contain too many highly technical terms that do not have synonyms. Most sentences are replete with terms that are BY DEFINITION not replaceable. Also, often the wording is already highly standardised and concise, which leaves me no room to paraphrase. 

When I try to explain this to my clients, they often fail to see the logic here and insist on me trying, which is a huge waste of their and my valuable time. I do understand the pressure faced by students who must submit their work and are confronted with a rigid system (Turnitin, etc) that cannot be argued with and offers no room for negotiation and compromise. In my opinion, such plagiarism software is only useful if it is used at the discretion of the thesis supervisors, but not as a mandatory hurdle at department and faculty level. 

However, this problem of reducing the similarity score to below 20 percent is solvable in other subject areas such as history, business, economics, education, philosophy, and the like. Although the professional level of these texts is equally demanding, their used terminology leaves more room for paraphrasing, and many terms can indeed be replaced.

Another challenge for me as the line editor is to change the syntax structure of each sentence, which is much easier to do when the sentences are rich and varied. I am able to change from active to passive voice, leave small details out or add them according to the context, and can thus change the individual building blocks and order of elements in a sentence enough to avoid similarity without losing the meaning. 

In a recent project I was able to reduce the similarity score from 43 percent to 10 percent, and from 64 percent to 16 percent. I managed to reduce the third chapter from 50 percent to 35 percent, but then hit a wall because the last chapter contained too many elements taken from previous studies that could not be replaced and had to be left as they were (measurement items, calculations, etc). 

In short, you can expect your editor to resolve this pesky issue of similarity, but bear in mind that there ARE limits to this kind of work. Your text has to allow paraphrasing and variation; if it is too scientific and formulated you will have to remove some of your sourced material...