知識地圖的建構方式及其內容與結構之真確性保護機制

Abstract

知識地圖有助於提升存取知識資訊的效率與效能，故本論文以階層成長式自組織映射圖技術為基礎，針對大量的數位文件資料，提出知識地圖可操作性的建構方法，並以兩項實作案例探討其可用性及可行性，分別為(1)建構Bill Gates言論資料的知識地圖，與(2)針對資訊安全教學研究導向的網站資料，建構其知識地圖。在第(1)項實作中主要是從個別文件中自動萃取關鍵字，而第(2)項則必須仰賴文件建立者自行輸入關鍵字。此兩項實作案例分別建構出不同特性的知識地圖，引導使用者從顯性知識中挖掘隱性知識的存在。 知識地圖下的知識內容及其地圖結構的真確性會影響所取用知識之正確與否。在此，真確性的意涵是指知識地圖之內容與結構沒有遭受未經授權的更改，因此，本論文提出一種真確性檢查機制來確保知識地圖的真確性。此機制將知識地圖下所有的知識內容視為一複合式文件，而此複合式文件是由大量的個別文件所組成，個別文件可以是不能分割的單元文件，也可以是由單元文件所構成。本論文所提出的機制是以碰撞阻抗赫序函數計算每個單元文件的真確性檢查碼，再進一步依照地圖與文件間的結構關係，產製整份複合式文件的檢查碼。此複合式文件的檢查碼可以讓檢查者確認個別文件及其順序結構之真確性，沒有遭到未經授權的變更。 綜合上述之研究內容，本論文是針對大量的數位文件資料，提出知識地圖建構與其真確性保護間相互關聯的解決方案。

Knowledge maps provide essential help for effectively and efficiently searching information from knowledge content, in particular, from a large amount of collections. This thesis applies the technique of Growing Hierarchical Self-Organizing Map (GHSOM) to propose a process for constructing a knowledge map from knowledge content. The content here is a collection of digital documents. Two implementations have been conducted: (1) constructing a knowledge map from a collection of Bill Gates’ speeches, and (2) constructing a knowledge map from the documents accumulated in a Web portal serving a teaching and research purpose at NCTU. The first implementation includes a task that automatically extracts the key words form each component document. The second implementation does not include such a task, requesting the document creator himself provides the key words. The two implementations create two knowledge maps, each of which yields different implications that guide user to catch certain implicit knowledge. The integrity of the knowledge content under the map and that of the map itself is a key factor for the correctness of searched information. The integrity requirement here is to ensure that the content and the map are not unauthorized modified. This thesis proposes an integrity inspection mechanism to ensure this property. The mechanism views the whole knowledge content under the map as a compound document that consists of a collection of component documents, wherein each component document is either a compound document as defined or a unit document that cannot be further decomposed. The mechanism applies a collision-resistant hash function to each unit document to obtain an integrity inspection code for the unit document. The mechanism then produces an integrity code for a compound document from the inspection codes of its component documents. The inspection code of a compound document allows an inspector to check whether the sequential structure and the content of its component documents have been unauthorized changed. As a result, this thesis provides inter-related solutions for constructing knowledge maps and inspecting integrity on a collection of digital documents.