【目的】 探讨学术不端系统查重过程中的人机协作及优化策略。 【方法】 采用AMLC、PSDS检测医学论文,筛选查重系统中可有效提示学术不端的参数及报告,以明确人机协作任务及优化流程。 【结果】 AMLC复制比存在假超限现象,但其全文对照报告提示的数据重复,对判定低复制比的研究型论文学术不端有较大参考价值;PSDS全文报告及参数未能提供有效参考。故人机协作的任务在于排除误判及查找低复制比的学术不端论文。AMLC检测提示复制比超限时,可快速浏览全文(标明引文)报告以排除假超限;对于复制比正常的研究型论文,需核查AMLC全文对照报告提示的数据重复。 【结论】 分析查重系统的提示特点,筛查有效提示学术不端的参数,可明确人机协工作任务,指导、优化人机协作。
[Purposes] The paper aims to explore man-machine coordination and optimized strategy in duplicate checking. [Methods] The systems of Academic Misconduct Literature Check (AMLC) and Paper Similarity Detection Service (PSDS) were used to detect medical manuscripts, from which parameters and reports that can effectively indicate academic misconduct were screened, so as to clarify the task of man-machine coordination and optimize its procedure. [Findings] AMLC overestimated the copy ratio of several manuscripts, while its data duplication revealed by full-text comparison report was valuable for determining academic misconduct in research papers with normal copy ratios. Parameters and reports by PSDS failed to provide valuable reference. Therefore, the task of man-machine coordination was to eliminate overestimation of the copy ratio and identify manuscripts with normal copy ratios. For AMLC detection, we can glance over the full-text (citations labeled) report to exclude those manuscripts with overestimated copy ratio, and inspect data duplication in the AMLC full-text comparison report for research papers with normal copy ratios. [Conclusions] Analyzing the features of detection systems to select valuable parameters and reports can clarify the task of man-machine coordination and optimize its procedure in academic misconduct detection.
Low copy ratio,