臺灣博碩士論文加值系統

本論文針對電子郵件分類與重要性預測的問題，提出了一個人化電子郵件助理系統：助理系統會學習使用者的分類習慣以及重要性歸類方法。並以此系統為基礎，進行分類準確率與重要性預測正確率的實驗。
本系統的核心是信件分類模組，利用在文件分類上有良好效果的 naive Bayes classifier 來實作。同時將信件表述成許多種不同的方式，以測試何種方式可以使郵件分類的正確率最高。
在重要性預測方面，如果將郵件的重要性分為三種，高、中、低，那麼郵件的重要性預測問題也可以利用分類問題來解決，因此對此我們也進行與分類問題同樣的實驗。針對網際網路上目前氾濫的廣告信件，我們也測試了系統核心過濾廣告信件的能力。
總之，經過一段時間的學習後，電子郵件助理便可以提供給個人越來越貼心的服務。

This thesis discusses the problem of email classification and prioritization. An email agent system is implemented. It can learn
a user''s customs for email classification. Based on this system, experiments of classification correctness and importance prediction
are done.
The kernal of this system is mail classification module. The naive Bayes classifier, which performs well on text classification is applied. Besides, we represent emails as several different forms to test which form is the best for improving correctness of classification .
In the aspect of importance prediction, if we divide the importance of emails into three levels, high, midium, and low, the problem can reduce to classificationo problem. As a result, we do the same experiments as classification problem. We also use our system to test the ability of jumk mail filtering.
In summary, after learning time, email agent can provide users better and better services.

CHAPTER 1 INTRODUCTION1
1.1PROBLEM2
1.2THESIS OVERVIEW5
CHAPTER 2 RELATED WORK9
2.1PERSONALIZED INFORMATION FILTERING9
2.2INTERFACE AGENTS13
2.3PRESENT EMAIL AGENTS15
2.4SUMMARY18
CHAPTER 3 MACHINE LEARNING FOR EMAIL CLASSIFICATION21
3.1DEFINITION OF THE NAIVE BAYES CLASSIFIER22
3.2REPRESENTATION OF A MESSAGE23
3.3CLASSIFICATION ALGORITHM26
CHAPTER 4 EXPERIMENTAL EVALUATION FOR CLASSIFICATION33
4.1CLASSIFIER IMPLEMENTATION33
4.2EXPERIMENTAL EVALUATION37
CHAPTER 5 PRIORITIZATION43
5.1CONCEPT AND IMPLEMENTATION43
5.2EXPERIMENTAL EVALUATION FOR PRIORITIZATION44
5.3EXPERIMENTAL EVALUATION FOR SPAM FILTERING45
CHAPTER 6 DEMONSTRATION49
6.1INSTALLATION49
6.2MAILSERVER PROPERTIES SETTING50
6.3HOW TO USE MAILKEEPER？51
CHAPTER 7 CONCLUSION AND FUTURE WORK55
7.1CONCLUSION55
7.2FUTURE WORK56
APPENDIX A DISCUSSION OF EMAIL AGENT SYSTEM ARCHITECTURE59
A.1MULTI-USER MAILKEEPER59
A.2GATEWAY60
A.2.1MAIL RETRIEVE AND PROCESS MODULE61
A.2.2MESSAGE DELIVER62
A.3CLIENT62
A.4COMMUNICATION PROTOCOL63
A.5DATA FLOW66
APPENDIX B CONDITIONAL PROBABILITY69
BIBLIOGRAPHY73

Bibliography
[Belkin & Croft, 1992] Belkin, N. J., & Croft, W. B. (1992). Information filtering and information retrieval: two sides of the same coin? Communication of the ACM, 35, 29-38.
[Boone, 1998] Boone, G. (1998, May 10-13). Concept features in Re:Agent, an intelligent email agent. Paper presented at the Second International Conference on Autonomous Agents, Minneapolis/St. Paul.
[Chien, 1997] Chien, L.-F. (1997). PAT-Tree-Based keyword extraction for Chinese information retrieval. Paper presented at the ACM SIGIR Conference, Philadelphia, USA.
[Cohen, 1996] Cohen, W. W. (1996). Learning trees and rules with set-valued features. Paper presented at the Thirteenth National Conference on Artificial Intelligence, Portland, Oregon.
[Cranor, 1998] Cranor, L. F., & LaMacchia, B. A. (1998). Spam! Communications of the ACM, 41, 74-83.
[Crocker, 1982] Crocker, D. H. (1982). Standard for the format of ARPA internet text messages (RFC 822) .
[Deng, 1998] Deng, Y.-H. (1998). Personalized E-Mail agent. Master''s Thesis, Dept. of Computer Science and Information Engineering, National Taiwan University.
[Dorner, 1992] Dorner, S. (1992). Eudora Reference Manual: Qualcomm Inc.
[Hermans & Schlimmer, 1993] Hermans, K. & Schlimmer, J. (1993). A machine learning apprentice for the completion of repetitive forms. Paper presented at the Ninth IEEE Conference on Artificial Intelligence for Applications, Orlando, Florida.
[Hsu & Yih, 1997] Hsu, J. Y.-J., & Yih, W.-T. (1997). Template-based information mining from HTML documents. Paper presented at the AAAI-97, Providence, USA.
[Joachims, 1996] Joachims, T. (1996). A probabilistic analysis of the Rocchio algorithm with TFIDF for text categorization (CMU-CS-96-118): Carnegie Mellon University.
[Kozierok & Maes] Kozierok, R., & Maes, P. (1993). A learning interface agent for scheduling meetings. Paper presented at the ACM SIGCHI International Workshop on Intelligent User Interfaces, Orlando, Florida.
[Lacey, 1997] Lacey, M. (1997). Conditional Probability. Unpublished Course Page,http://www.stat.yale.edu/Courses/101/condprob.htm.
[Lai et al., 1988] Lai, K., Malone, T., & K., Y. (1988). Object lens: A spreadsheet for cooperative work. ACM Transactions on Office-Information Systems, 5, 297-326.
[Lang, 1995] Lang, K. (1995). NewsWeeder: Learning to filter netnews. Paper presented at the the Twelfth International Conference on Machine Learning, San Francisco, CA.
[Lashkari et al., 1994] Lashkari, Y., Metral, M., & Maes, P. (1994, August 1994). Collaborative interface agents. Paper presented at the Twelfth National Conference on Artificial Intelligence, Seattle, WA.
[Lewis, 1991] Lewis, D. (1991). Representation and learning in information retrieval, (Ph.D. thesis), (COINS Technical Report 91-93). Dept. of Computer and Information Science, University of Massachusetts.
[Lieberman, 1994] Lieberman, H. (1994). Attaching interface agent software to applications .
[Maes & Kozierok, 1993] Maes, P., & Kozierok, R. (1993). Learning interface agents. Paper presented at the Eleventh National Conference on Artificial Intelligence, Washington D.C.
[Maes, 1994] Maes, P. (1994). Agents that reduce work and information overload. In J. M. Bradshaw (Ed.), Software Agents (pp. 145-164): AAAI Press/The MIT Press.
[McCallum & Nigam, 1998] McCallum, A., & Nigam, K. (1998). A comparison of event models for naive Bayes text classification. Paper presented at the AAAI-98.
[Mitchell, 1997] Mitchell, T. M. (1997). Machine Learning: McGraw-Hill.
[Mladenic & Grobelnik, 1998] Mladenic, D., & Grobelnik, M. (1998, April 21 - 24). Efficient text categorization. Paper presented at the Text Mining Workshop on the 10th European Conference on Machine Learning, Dorint-Parkhotel, Chemnitz, Germany
.
[Myers, 1996] Myers, J. (1996). Post Office Protocol - Version 3 (RFC 1939) .
[Payne & Edwards] Payne, T., & Edwards, P. (1997). Interface agents that learn: An investigation of learning issues in a mail agent interface. Applied Artificial Intelligence, 11, 1-32.
[Pazzani et al., 1996] Pazzani, M., Muramatsu, J., & Billsus, D. (1996). Syskill & Webert: Identifying interesting web sites. Paper presented at the National Conference on Artificial Intelligence, Portland, OR.
[Quinlan, 1986] Quinlan, J. R. (1986). Induction of decision trees. Machine Learning, 1(1), 81-106.
[Rennie, 1998] Rennie, J. (1998). ifile: An application of machine learning to mail filtering .
http://www.cs.cmu.edu/~jr6b/papers/ifile98.ps.gz
[Russell & Norvig, 1995] Russell, S., & Norvig, P. (1995). Artificial intelligence: A modern approach. Englewood Cliffs, NJ: Prentice-Hall.
[Salton, 1989] Salton, G. (1989). Automatic Text Processing: The Transformation, Analysis, and Retrieval of Information by Computer: Addison Wesley.
[Scott, 1998] Scott, S. (1998). Feature engineering for a symbolic approach to text classification. Master''s Thesis, Dept. Computer Science, University of Ottawa.
[Sheth, 1994] Sheth, B. (1994). A learning approach to personalized information filtering. Master''s Thesis, Dept. of Electronic Engineering and Computer Science, MIT.
[Stanfill, 1986] Stanfill, C., & Waltz, D. (1986). Toward memory-based reasoning. Communication of the ACM, 29, 1213-1228.