頁籤選單縮合
題 名 | 利用網路結構分析的研究主題視覺化=A Study of Research Topic Visualization Using Network Structure Analysis |
---|---|
作 者 | 林頌堅; | 書刊名 | 教育資料與圖書館學 |
卷 期 | 50:4 2013.夏[民102.夏] |
頁 次 | 頁565-596 |
分類號 | 028 |
關鍵詞 | 研究主題分析; 資訊視覺化; 路徑搜尋網路; 社群偵測演算法; Analysis of research topics; Information visualization; Path-finder networks; Community detection algorithms; |
語 文 | 中文(Chinese) |
中文摘要 | 路徑搜尋網路(PFNet)方法與社群偵測演算法經常被應用在研究主題的視覺化呈現與分析。PFNet方法能夠在保留原先網路的結構特性下,刪除大量不重要的連結線。社群偵測演算法則能夠將網路劃分成凝聚性子群。然而這兩種方法都有不足的地方:PFNet方法無法自動從輸入的網路上發現重要的子群,社群偵測演算法無法保證同一子群的節點會映射在鄰近的區域。本論文建議整合這兩種方法以減輕上述的問題:利用社群偵測演算法將PFNet方法產生的新網路劃分成子群。並且本研究也建議利用子群內的出現頻率較高的詞語做為研究主題的標示,讓結果分析與詮釋更加容易。本研究以臺灣資訊傳播學領域為範例,利用相關系所的碩士論文為分析資料。研究結果發現:整合PFNet方法和社群偵測演算法有利於從論文相關網路上發現代表重要研究主題的子群。子群內最高出現頻次的詞語大多和資訊傳播學以及其基礎領域的問題、方法、理論和技術非常相關,可以做為研究主題的標示。 |
英文摘要 | The Pathfinder network (PFNet) method and the community detection algorithms both are methods which have been widely applied to visual presentation and analysis of research topics. The PFNet method can delete a large amount of insignificant links in networks but also retains the structural characteristics of the original networks, while the community detection algorithms are able to partition networks into a set of cohesive subgroups. However, each of the methods has its deficiencies. The PFNet method cannot automatically find out critical subgroups in input networks and the community detection algorithms do not guarantee nodes in the same subgroup able to be mapped in neighboring area. The integration of these two methods provides a way to alleviate the above problems: The output network from the PFNet method is partitioned using community detection algorithm. In addition, this study also suggests that the use of high frequency terms within papers in subgroups as the labels of research topics to make the analysis and interpretation of results easier. This study takes the field of Information Communication as an analytic case to study the application of the integrated methods, and the data of master theses of the related graduated schools are collected to be used for the analysis. The results show that it is effective to integrate the PFNet method and the community detection algorithms to discover subgroups representing important research topics from the network which is constructed upon the relations between papers in the examined field. The terms with high occurring frequency in subgroups are very relevant to the problems, method, theorems and technologies in the field of Information Communications and its fundamental disciplines, and therefore, they are suitable as the representatives of research topics. |
本系統中英文摘要資訊取自各篇刊載內容。