Background

Scientists communicate progress and exchange information via publication and presentation at scientific meetings. We previously showed that text similarity analysis applied to Medline can identify and quantify plagiarism and duplicate publications in peer-reviewed biomedical journals. In the present study, we applied the same analysis to a large sample of conference abstracts.

Methods

We downloaded 144,149 abstracts from 207 national and international meetings of 63 biomedical conferences. Pairwise comparisons were made using eTBLAST: a text similarity engine. A domain expert then reviewed random samples of highly similar abstracts (1500 total) to estimate the extent of text overlap and possible plagiarism.

Results

Our main findings indicate that the vast majority of textual overlap occurred within the same meeting (2%) and between meetings of the same conference (3%), both of which were significantly higher than instances of plagiarism, which occurred in less than .5% of abstracts.

Conclusions

This analysis indicates that textual overlap in abstracts of papers presented at scientific meetings is one-tenth that of peer-reviewed publications, yet the plagiarism rate is approximately the same as previously measured in peer-reviewed publications. This latter finding underscores a need for monitoring scientific meeting submissions – as is now done when submitting manuscripts to peer-reviewed journals – to improve the integrity of scientific communications.

中文翻译：

---

估计生物医学会议摘要中文本重叠的普遍性

背景

科学家通过在科学会议上发表论文和发表论文来交流进展并交换信息。我们之前已经证明，应用于Medline的文本相似性分析可以识别和量化窃，并在同行评审的生物医学期刊中重复发表论文。在本研究中，我们将相同的分析应用于大量会议摘要中。

方法

我们从63个生物医学会议的207个国家和国际会议中下载了144,149个摘要。使用eTBLAST：文本相似性引擎进行成对比较。然后，领域专家审阅了高度相似的摘要（共1500个）的随机样本，以估计文本重叠的程度和可能的窃。

结果

我们的主要发现表明，绝大多数文字重叠发生在同一会议（2％）之内和同一会议的两次会议之间（3％），两者均明显高于抄袭情况，抄袭发生率不到0.5。摘要的百分比。

结论

该分析表明，科学会议上发表的论文摘要的文本重叠是同行评审出版物的十分之一，但，窃率与先前在同行评审出版物中测得的rate窃率大致相同。后一个发现强调需要监视科学会议的提交情况（就像在将稿件提交给同行审阅的期刊时所做的那样），以提高科学交流的完整性。