臺灣博碩士論文加值系統

由於資訊科技的發達，資料與訊息已被廣泛的產生與散佈，而就有釵h資訊系統是專門管理與分類這些大量的資料跟訊息。我們發現，有些工作是可以用文字描述的，那麼我們就可以根據這些描述的文字來分派這些工作，這個概念可以應用在分發公文上。而且如果我們使用文件分類的技術應用在分發公文(工作分派)上，那麼就可以達到自動或半自動的效果。工作分派自動化不只是能節省人力，人員職位更換時工作交接也不會有相關的問題發生。考量到分發的效能，我們是利用統計的方法來當作分類的方法，主要就是根據字彙出現的次數當作權重，且加入連續字權重加乘的計算方式，算出文件屬於各個類別的可能性。從我們的模擬數據可發現，即使系統的知識庫裡面已有40000筆文件，分類一份文件的整個過程不花超過20毫秒，而且分類的正確率也能高達七成。

Some systems are built to manage those information, like information
classification which differentiates between the useful and the
useless or classifies information into categories. Currently, we
found that some jobs are described into words, so we can assign jobs
according to the words. This concept could be used in dispatching
official documents. If we use techniques of document classification,
dispatching official documents (job assignments) would be done
automatically or semi-artificially. Automatic job-assignment could
not only save human effort but maintain the knowledge and keep the
precision of the assignment. Considering performance and efficiency
of the system, we use the statistical method of document
classification to calculate the possibility of each category by term
frequencies as term weights and continuous terms for increasing
weights. In our simulation, the whole classification procedure costs
less than 20 ms even if there are 40000 documents in the domain
knowledge, and the precision of our method could reach 70%.

1 Introduction
2 Related Works
2.1 Term Weighting
2.2 Data Mining
2.3 Information Retrieval
2.4 Document Classification
2.5 Machine Learning
3 Motivation
4 Our Method
4.1 Method Description
4.1.1 Classification
4.1.2 Learning
4.1.3 Example
4.2 System Architecture
5 Experiment
5.1 Experimental Results
5.2 Current Problems
6 Future Work
7 Conclusion

C. Deerwester and Susan T. Dumais and Thomas K. Landauer and George W.
Furnas and Richard A. Harshman, "Indexing by Latent Semantic Analysis." Journal
of the American Society of Information Science, pp. 391-407, 1990.
[3] N. Cristianini and J.Shawe-Taylor, An Introduction to Support Vector Machines. Cam-
bridge University Press, 2000.
[4] Vladmimir N. Vapnik, the Nature of Statistical learning Theory. Springer, 1995.
[5] Joachims, T. (1998), "Text categorization with support vector machines: learning with
many relevant features." Proceedings of ECML-98, 137-142.
[6] Belur V. Dasarathy, Nearest Neighbor Norms: NN Pattern Classi¯cation Techniques,
Los Alamitos, CA: IEEE Computer Society Press, 1991.
[7] Shakhnarovish, Darrell, and Indyk, Nearest-Neighbor Methods in Learning and Vision,
The MIT Press, 2005.
[8] http://en.wikipedia.org/wiki/Data mining
[9] W. Frawley and G. Piatetsky-Shapiro and C. Matheus. "Knowledge Discovery in
Databases: An Overview.", AI Magazine, Fall 1992, pp. 213-228.
[10] D. Hand, H. Mannila, P. Smyth, Principles of Data Mining. MIT Press, Cambridge,
MA, 2001.
30
[11] Hearst, M. What is text mining. http://www.sims.berkeley.edu/ hearst/text-mining.html,
(2004).
[12] W. Fan, L. Wallace, S. Rich, Z. Zhang. "Tapping into the power of text mining",
Communications of ACM, forthcoming, 2005.
[13] C. J. van RIJSBERGEN B.Sc., Dip. NAAC, Ph.D., M.B.C.S., F.I.E.E., C.Eng.,
F.R.S.E. INFORMATION RETRIEVAL. Online book on
http://www.dcs.gla.ac.uk/Keith/Preface.html
[14] http://en.wikipedia.org/wiki/Information retrieval
[15] Marti A. Hearst, "Untangling text data mining.", In Proceedings of the 37th conference
on Association for Computational Linguistics, pages 3{10, College Park, Maryland,
1999. Association for Computational Linguistics.
[16] Yang, y. (1990), "An evaluation of statistical approaches to text categorization.",
Journal of Information Retrieval, 67-88.
[17] Aggarwal, C. C., and Yu, P. H(2001), "On e®ective conceptual indexing and similarity
search in text data." Proceedings of the 2001 IEEE International Conference on Data
Mining (pp. 3-10). San Jose.
[18] S. R. Safavin and D. Langrebe, "A survey of decision tree classi¯er methodology.",
IEEE Transactions on Systems, Man and Cybernetics, 21(3):660{674, 1991.
[19] Sung-Shun Weng and Chi-Kai Liu, "Using text classi¯cation and multiple concepts to
answer e-mails.", Expert Systems with Applications, Vol. 26, No. 4, pp. 529-543. (SCI,
EI).
[20] Boser, B.E., Guyon, I. M., and Vapnik, V. , "A Training Algorithm for Optimal Margin
Classi¯ers.", Fifth Annual Workshop on Computational Learning Theory, ACM, 1992
[21] H. Yu, J. Han, and KC-C. Chang, "PEBL: Web Page Classi¯cation without Negative
Examples." IEEE Trans. Knowledge and Data Engineering, vol. 16, 2004.
31
[22] Chih-Chung Chang and Chih-Jen Lin. LIBSVM. http://www.csie.ntu.edu.tw/ cjlin/libsvm/
[23] http://www.w3.org/XML/