engleski

The frequency of scientific plagiarism measured by a text matching software: A systematic review

Today, computer programs easily detect textual similarity between texts which enables us to measure frequency of plagiarism in scientific articles. Data about frequency and characteristics of scientific plagiarism incite discussions about more precise definition of scientific plagiarism and about possible ways to prevent and regulate plagiarists in specific institution or journal. The aim of this study is to assess the frequency and characteristics of plagiarism in the international scientific community with purpose to incite discussion about universal definition of plagiarism and measures to prevent it. For this purpose we conducted a systematic review of published scientific studies that detect the frequency of plagiarism in scientific articles using text matching software. We conducted literature search which focus on articles about research ethics, with the aim of finding potentially relevant studies. We included studies that measured frequency of plagiarism on a sample of scientific papers that had met at least one of the following three conditions: papers submitted for publication in a scientific journal, or papers accepted at the scientific conference, or papers published and indexed in bibliographic databases of scientific papers. We excluded studies that do not measure the frequency of plagiarism or do not present results as an absolute (n) and relative (%) frequency or measure the frequency of plagiarism on a sample of students' term papers, theses, master's theses and doctoral dissertations. Following data were collected from included studies: basic characteristics of a study (the first author, year of publication, period, country), characteristics of analyzed research papers (number, type and a status of analyzed paper), basic characteristics of a procedure for plagiarism detection (name of text-matching tool, names of databases or other sources of scientific content with whose content is compared analyzed work, criteria used to identify plagiarism) and outcome measures: number of analyzed scientific papers that contain a mild (technical) form of plagiarism and the number of analyzed papers that contain serious form of plagiarism, the total number of analyzed works with any form of plagiarism. Two types of studies are identified: studies with large sample (more than 60, 000 papers analyzed) that identify plagiarism primarily based on the results of text-matching tool, and studies with small sample (less than 2500 analyzed works) that identify plagiarism based on the results of text-matching tool and results from a method of comparative reading of two or more similar texts. Studies with large samples of analyzed papers (N = 3) estimates that 0, 2% to 1% pair of papers contain possible plagiarism, while studies with small sample (N = 4) estimate that 5% to 23% of the analyzed papers contain a possible plagiarism. It was also found that the milder forms of plagiarism, e.g., copy of a standard method or technical terms, are more frequent than serious forms of plagiarism e.g., copy of ideas or research results.

plagiarism ; scientific misconduct ; text-matching software ; research integrity

nije evidentirano