Fig. 1
From: Estimating the prevalence of text overlap in biomedical conference abstracts
Summary of eTBLAST and expert classification. Our study was conducted from left to right. Far left, conference proceedings were web-scraped or pdf-scraped and deposited into a free online database: http://www.ethicsdb.org/. We considered 207 meetings of 63 different conferences. Middle left, we used eTBLAST to compare abstracts within each conference meeting (gray box), between each conference (red box), and to published abstracts in medline (blue box). Middle right, random samples of highly similar abstracts identified by eTBLAST were evaluated for misconduct by a domain expert (ethicist). Far right, random sampling and classification identified 126 (12.6%) instances of text overlap – same meeting, 223 (22.3%) instances of text overlap – same conference, and 238 (23.8%) instances of text overlap with MEDLINE. Bone-fide plagiarism was rare; we identified 9 potential instances. We concluded that text overlap – same meeting, text overlap – same conference, and plagiarized abstracts account for approximately 2, 3%, and .5% of conferences abstracts, respectively