臺灣博碩士論文加值系統

網際網路是現今取得資訊最佳的管道。但是在面對這包羅萬象的資料庫時，最大的問題在於如何有效的取得所需的資訊。搜尋引擎是一個不錯的工具。只要我們輸入一些關鍵字，它能協助我們迅速的找到相關的資料。但是搜尋引擎並非萬能。至少在一些特定的情況下，它無法有效的協助使用者快速的找到資料。致使我們常需要瀏覽上百篇的網頁敘述，運氣好才能找到一篇有用的資訊。另外，搜尋引擎無法將搜尋結果分類，造成我們為了要找一些主題相關的文章時，需要花釵h時間一頁一頁的瀏覽。最常被舉出的例子是當我們想知道美洲馬鴝陳銃]多快，於是我們輸入一組關鍵字“speed of jaguar”。結果，我們得到數以萬計關於汽車的文章，卻需要瀏覽數百篇敘述，才能找到一篇與美洲隻傢鰝滌T息。這是目前搜尋技術所無法突破的障礙，也是我們想利用資訊視覺化(information visualization)技術來加強的它的原因。
這篇論文主旨在介紹我們發展出來的VisSearch 系統。VisSearch 的名子是取自‘visualize the result of search’。該系統應用的釵h資訊視覺化與資訊擷取(information retrieval)的理論，加上我們自己的一些研究心得所完成。系統的基本假設是：
1. 一篇文件的主題可以被少數幾個關鍵字眼所闡示。
2. 關鍵字可以從字眼出現在一篇文章中的位址與頻繁度來決定。
3. 我們可以憑藉著展示文章與關鍵字間的對應關係，來視覺化搜尋結果。
這三個基本假設的可行性將隨著VisSearch 系統的開發而被驗證。完成後的系統將能夠協助使用者分類、標示與快速得知該次搜尋的結果。它就像一張地圖般的協助使用指找到大類方向，或是讓使用者知道需要重選關鍵字進行下一次搜尋，以期取得更精確的結果。
論文先介紹一些關於資訊視覺化與資訊擷取的理論，但是僅就與本系統相關的部分說明。接下來我們會說明VisSearch 系統的設計細節，並會展示該系統實際應用例子，讓讀者感受該系統如何有效的協助使用者尋找所需的資訊。然後，最後再討論現存的問題與一些可以改善系統弁鄋漲a方。
我們期望VisSearch 系統是一個簡單好用的工具，沒有太複雜的介面，輸出結果淺顯易懂，並且真正能有效的幫助使用者解決問題。

Recently, Internet has become the best way for retrieving information. But how we can efficiently retrieve the information is the major problem when we face this enormous database. Search engine provides a good means for this. Just input several keywords, Search engine will find relative documents for us rapidly. But search engine is not perfect. At least in some situations, search engine can’t find related documents precisely. Therefore the user has to spend much time to browse the result page by page. In such situation, finding a useful document would depend on lock. The most popular example is that when we would like to know how fast a jaguar could run. For this purpose, we choose “speed of jaguar” as a keyword to search. As a result, we got thousands of document related to automotive and car racing. We just can find one document related to our purpose only after reading hundreds of descriptions. This is the major obstacle of retrieval technology currently used. The obstacle motivates us to enhance it by applying the technology of information visualization.
The subject of this thesis is to introduce our VisSearch system. The name of system stands for ‘visualize the result of search’. The system is built upon applying numerous theories of information visualization and information retrieval, plus some assumptions. The basic assumptions of the system is:
1. Document can be denoted by some significant words.
2. Significant word can be identified by its occurrence and frequency.
3. Search results can be visualized by showing relationship between the significant words and the documents
The feasibility of three assumptions will be proved by the implementation of VisSearch system. The result of implementation will be able to assist user in classifying, labeling so as to approach their target efficiently. It can be used as a map to help users find a direction to relative documents, or let users know how to change the search keywords to retrieve more accurate results.
In the content, first we will introduce several theories of information visualization and information retrieval. The scope is restricted to theories that have been adopted in our system. Next, we will introduce the VisSearch system and demonstrate how the system help user when searching. After that, we will discuss current issues of the system and possibilities of enhancement and future works.
When designing VisSearch system, we intent to make the system easy to use, without a complex interface, and the output of the system is intuitive and comprehensible. Ultimately, we hope the system will actually help users to solve their problems.

Chapter 1 Introduction 1
1.1 Problem definition 2
1.2 Review of previous works 3
1.3 Scope and direction 5
Chapter 2 Background 7
2.1 Theory of information visualization 7
2.1.1 Introduce of information visualization 7
2.1.2 Essences of information visualization 8
2.1.3 Examples of information visualization 12
2.2 Theory of information retrieval 15
2.2.1 Introduce of information retrieval 15
2.2.2 Vector model of retrieval 15
2.2.3 Variation of vector model 17
Chapter 3 Implementation 21
3.1 System Overview 21
3.1.1 Basic assumptions 21
3.1.2 Presentation 22
3.1.3 Representation 24
3.1.4 Interaction 25
3.2 Implementation Details 25
3.2.1 Document preprocess 25
3.2.2 Analyzing relationships among terms 25
3.2.3 Plotting documents 27
3.2.4 Labeling 30
3.2.5 Architecture for extensibility 30
Chapter 4 Demonstration 32
4.1 Basic instance 32
4.2 Second instance – Large information items 40
4.3 Third instance – The ‘jaguar’ problem 43
Chapter 5 Discussion 47
5.1 Problems about term’s relationship analyze 47
5.2 Problems about presentation 47
5.3 Problems about stemming 49
5.4 Influence of parameters 49
5.5 Increasing availability 49
Chapter 6 Conclusion and Future work 51
Postscript 53
Bibliography 54
Appendix 56

[Bouvier1999]
D. Bouvier. Getting Started with the Java 3D API – A tutorial for beginners. Sun press, 1999.
[Eaton2002]
C. Eaton and H. Zhao. Visualizing Web Search Results. 2002
[Frakes1992]
W. Frakes and B. Ricardo. Information retrieval: Data structures and algorithms. Prentice Hall press, 1992.
[Google2003]
Google’s home page- http://www.google.com/
[Humphreys1989]
G. Humphreys and V. Bruce. Visual Cognition 1989.
[J3D2003]
J3D’s home page- http://www.j3d.org
[JAVA2003]
Java’s home page- http://java.sun.com/
[Mackinlay1991] J. Mackinlay, G. Robertson and S. Card. The perspective wall: detail and context smoothly integrated Mackinlay color plates. 1991
[McCrickard1997]
D. McCrickard and C. Kehoe. Visualizing Search Result using SQWID, 1997.
[Olsen1991]
K. Olson, R. Kofhage, K. Sochats, B. Spring and J. Williams. Visualization of a document collection: theVIBE system. 1991
[Oren1998]
Z. Oren and E. Oren. Web Document Cluster, A flexibility demonstration. ACM 1998.
[Porter1980]
M. Porter’s personal web- http://www.tartarus.org/~martin/PorterStemmer/.
[Ricardo 1999]
B. Ricardo and R. Berthier. Modern Information Retrieval. ACM press, 1999.
[Sowizarl 2000]
H. Sowizarl, K. Rushforth and M. Deering. The Java 3D API Specification, Second Edition. Sun press, 2000.
[Steinbach2000]
M. Steinbach, G. Karypis and V. Kumar. A Comparison of Document Clustering Techniques. 2000.
[Spence2001]
R. Spence. Information Visualization, ACM press, 2001.
[TRTC2003]
Taipei Rapid Transit Coporation’s web site- http://home.trtc.com.tw
[Ware2000]
C. Ware. Information Visualization: Perception for Design, 2000.