- 期刊
利用弦圖依據論文關鍵字分類期刊作者的研究主題
Using chord diagrams to classify research themes of author terms for a scholar journal based on article keywords
摘要
許多研究利用社會網絡，分別探索作者合作團隊或關鍵字的主題分類，卻未見其適當集群分析方法的應用。本研究利用作者研發出來的集群演算法（稱作主從關係集群演算法），配合和弦圖的繪製，依據論文關鍵字分類期刊作者的研究主題。自華藝圖書資料庫下載「醫療資訊雜誌」歷年605篇作者姓名及其論文關鍵字，分別進行社會網絡分析，挑出論文篇數最多的合作團隊及研究主題，以其集群內中心度數最高者為其命名代表。再依據合作團隊論文的關鍵字，對照主題分類，再用和弦圖呈現研究結果。研究結果，顯示作者研發出來的集群演算法，只取用跟隨者與領導者間的最大唯一連結，因此見不到關鍵字集群間的連結，卻精簡與美化了視覺圖。利用和弦圖顯示頻次最多的20個關鍵字間的集群現象，其領銜者分別為電子病歷、資訊模組、糖尿病、資料倉儲、語音辨識、及決策支援系統。本研究依據論文關鍵字分類期刊作者的研究主題，示範和弦圖能夠呈現集群間的關聯，提供未來組織內團隊合作可視圖分析主題分類之參考。
並列摘要
Many studies have used social networks to explore the classification of author collaboration teams or keywords, but appropriate cluster analysis methods have not been applied. This study uses a cluster algorithm developed by the authors (referred to as the follower-leading clustering algorithm, FLCA) and chord diagram drawing to classify research topics based on journal author keywords. From the "Taiwan Medical Information Journal" database of the Airiti Library, 605 author names and their paper keywords were downloaded and analyzed using the FLCA algorithm. The team (or theme) with the most papers was selected, and the highest degree of centrality within the cluster was used as the representative name. Based on the keywords of the team's papers, their topic classification was compared, and the research results were presented using chord diagrams. The results showed that the cluster algorithm developed by the authors only uses the maximum unique link between followers and leaders, so the links between keyword clusters cannot be seen. The chord diagram displays the clustering phenomenon of the top 20 keywords, with electronic medical records, information modules, diabetes, data warehousing, speech recognition, and decision support systems as the leading topics. Based on the classification of paper keywords, this study demonstrates how chord diagrams can display the relationships between clusters, providing a reference for visualizing team collaboration and topic classification within organizations in the future.