學術論文之知識演化趨勢及其可視化研究

Mining and Visualization of Knowledge Evolution From An Academic Archive

李庭嫣(Ting-Yen Lee)

指導教授 ： 陳炳宇

共同指導教授 ： 許永真(Jane Yung-jen Hsu)

國立臺灣大學/電機資訊學院/資訊網路與多媒體研究所/碩士(2010年)

https://doi.org/10.6342/NTU.2010.02206

摘要

隨著Web2.0的普及，網際網路上充滿各式各樣的知識。然而因為這麼龐大的知識庫的組織卻是零散的，造成了人類學習新知的困擾。因此，我們希望藉由自動化的提供各種知識的發展流程，當人類需要學習新知時，可以藉由這些新知的來源與原來自己的知識互相比對，進而了解相關領域的結構，幫助人類快速進入狀況。

在本論文中，我們聚焦於大多數學生最容易遇到的學術論文的知識發展趨勢。我們先將論文所描述的主題用關鍵字代表，結合論文間會相互引述的關係，可以將相關的主題整合在一起，並且找到這些相關主題之間發展的關係，如：合併、分裂等。
藉由提供知識演化趨勢的發展，人們可以在踏入新的領域前，先對此領域有個基礎的認知，幫助人類的學習。

我們也收集了32,044篇學術論文來驗證，結果顯示我們的方法可以有效的將相關主題整合，並且提供知識的演化發展。

關鍵字

學術論文 ； 知識演化 ； 趨勢 ； 可視化研究

並列摘要

The World Wide Web is a rich source of knowledge. As many documents on the web are unstructured, it is difficult for people to learn efficiently in a systematic fashion. To facilitate learning, we aim to organize web resources according to the evolution of topics, which is a common organization used in many textbooks. In this research, we propose an approach to mining knowledge evolution from an academic archive by taking advantage of the citations between the literatures. Relationships among keywords are first extracted from each paper, and a citation graph is constructed from the references among papers to capture their dependency over time. Similar keywords can then be clustered into aggregated topics to track the evolution of knowledge. Finally, we propose methods to visualize knowledge evaluation by viewing both the evolution and interaction of aggregated topics. The experiments are conducted on 32,044 papers crawled from the ACM Digital Library, and the results show that the proposed approach is effective in mining knowledge evolution to help users learn from online resources.

並列關鍵字

Mining ； Visualization ； Knowledge Evolution ； Academic Archive

參考文獻

[3] R. Atterer and M. Tafelmayer. Treemap-based website navigation for non-hierarchical, interlinked sites: the trackback map. In Proceedings of the 2008 Nordic

[4] R. Bade, S. Schlechtweg, and S. Miksch. Connecting time-oriented data and in-formation to a coherent interactive visualization. In ACM CHI 2004 Conference

[5] M. Balzer and O. Deussen. Exploring relations within software systems using treemap enhanced hierarchical graphs. pages 89–94, 2005.

[9] J. Boyd-Graber, J. Chang, S. Gerrish, C. Wang, and D. Blei. Reading Tea Leaves: How Humans Interpret Topic Models. Advances in Neural Information Processing

[10] C. Chen. CiteSpace II: Detecting and visualizing emerging trends and transient