作者論文合作提攜後進的指標：PubMed即時分析

Evaluating Senior Authors' Mentorship Through Research Collaboration: Real-Time in PubMed

學術貢獻之評估不宜僅以論文篇數作為依據，而應更重視研究品質與其在學術社群中的實際影響力或貢獻，其中論文被引用量為最常用的指標，尚未見作者論文合作提攜後進的貢獻力指標。多數學術成果仍停留於靜態的作者論文列表，難以即時反映作者合作的結構與貢獻、或是其學術主宰力的資訊。本研究（1）開發一套PubMed線上即時視覺圖的作者合作分析應用程式（App4PubMedinR），將作者提攜後進分為5個等級，透過瀏覽器即可檢視作者合作提攜後進的分類，也（2）能批次計算機構內每作者提攜後進的強度，並（3）驗證其假設4位醫學資訊領域學者提攜後進的指標，皆被分類為強級（≥0.70）。結果顯示，（1）作者合作分析應用程式提供百餘張視覺圖；（2）批次計算1，407位2025年前2％頂尖台灣學者的提攜後進指標，平均值為0.62，中位數為0.65，說明不是每學者的指標皆大於0.70；（3）4位醫學資訊領域學者提攜後進的分類，3人分類為很強，一人為稍強，指標分別為0.81、0.81、0.77、及0.69。本研究所示範提攜後進的分類指標，結合PubMed即時視覺圖分析，能有效辨識學者提攜後進之強度，可以提供作為學者或機構作者論文合作提攜後進貢獻指標建立的參考。

Academic contribution should not be evaluated solely by publication counts. Greater attention should also be given to research quality, scholarly influence, and contributions to academic collaboration. Although citation counts and related indicators are widely used, few metrics specifically capture an author’s role in mentoring junior researchers through collaborative publications. In addition, most academic outputs are presented as static publication lists, which limits real-time interpretation of collaboration structures, contribution patterns, and academic leadership.This study aimed to: (1) develop App4PubMedinR, a PubMed-based author collaboration analysis application that provides real-time visualizations and classifies mentorship strength into five levels through a web browser; (2) enable batch computation of mentorship indicators for authors within institutions; and (3) demonstrate the proposed indicator using four scholars in medical informatics.The application generated more than 100 types of collaboration-related visualizations. In a batch analysis of 1,407 Taiwanese scholars ranked among the global top 2% scientists before 2025, the mean mentorship indicator was 0.62 and the median was 0.65, indicating that not all highly ranked scholars reached the predefined strong threshold of 0.70. Among the four medical informatics scholars, three were classified as having very strong mentorship and one as having moderately strong mentorship, with indicator values of 0.81, 0.81, 0.77, and 0.69, respectively.The proposed mentorship classification indicator, combined with PubMed-based real-time visual analytics, may help characterize mentorship patterns in academic collaboration. It may also serve as a practical reference for developing contribution metrics related to mentoring junior researchers at individual and institutional levels