臺灣博碩士論文加值系統

本研究提出一個自動擷取電子論文書目資料（論文標題、作者、及出版資訊）之系統。系統所擷取到之書目資料為連貫的，並包含了對應其電子論文之標題、作者、及出版資訊。
欲解決論文書目資料擷取問題，本研究利用電子論文中所擷取出的字型資訊，及隱馬可夫模式（一個機器學習之統計模型）來輔助，並且運用線上電子論文書目索引系統對系統擷取資訊進行修正。本研究實施各種實驗以測試本研究所設計之系統表現出的正確率。其中包括本研究中設計的隱馬可夫模式模型在不同情況下的表現，及整個系統本身的正確率。整體而言，此系統在電子論文書目資料擷取之正確率達到不錯的表現。

This study describes an automatic information extraction system for extracting bibliographical information (i.e. title, author, and publication information) from digital academic literature/research paper/E-paper. The bibliography information found should be non-fragmented, consisting the title, author(s), and publication information of the corresponding extracted paper.
To solve bibliography extraction problem, this study makes use of font size information extracted from a paper and available online literature databases, with aid of Hidden Markov Model (a statistical machine learning model). To test the viability of the system and framework, experiments are conducted to evaluate the accuracy of the system designed in this study, including the performances of the HMM model in different circumstances and the entire performance of the system itself. Overall, the system yields a fair performance in extracting bibliography information.

Abstract ................................................................................................................................................................. I
摘要……… ............................................................................................................................................................. II
誌謝……….. .......................................................................................................................................................... III
Table of Contents ................................................................................................................................................. IV
List of Figures ....................................................................................................................................................... VI
List of Tables ....................................................................................................................................................... VII
1. Introduction ............................................................................................................................................ 1
1.1. Background ..................................................................................................................................................... 1
1.2. Thesis Organization ........................................................................................................................................ 4
1.3. Problem Space ................................................................................................................................................ 4
1.3.1. Problem Definition .......................................................................................................................................... 4
1.3.2. Information Extraction ................................................................................................................................... 4
1.3.3. Unstructured Text ........................................................................................................................................... 5
2. Related Work .......................................................................................................................................... 6
2.1. Rule-based Model ........................................................................................................................................... 6
2.1.1. Rule-based Extraction for Author Information ................................................................................................ 7
2.1.2. Rule-based Extraction for journal information ................................................................................................ 8
2.1.3. Rule-based Extraction for Title ........................................................................................................................ 9
2.2. Hidden Markov Model .................................................................................................................................. 10
2.2.1. Simple Example of HMM ............................................................................................................................... 10
2.2.2. Notations ...................................................................................................................................................... 10
2.2.3. Canonical Problems in HMM ......................................................................................................................... 11
2.2.4. HMM Visualization ....................................................................................................................................... 12
2.2.5. HMM assumptions ....................................................................................................................................... 13
2.2.6. HMM Model ................................................................................................................................................. 14
2.2.7. Smoothing methods ..................................................................................................................................... 15
2.2.8. Viterbi Algorithm .......................................................................................................................................... 15
2.2.9. Different HMM state transition model .......................................................................................................... 16
2.3. Review of Previous Studies ............................................................................................................................ 17
3. Proposed Methods ................................................................................................................................ 20
3.1. System Architecture ...................................................................................................................................... 20
3.2. Input Subsystem ........................................................................................................................................... 20
3.3. Extraction Engine .......................................................................................................................................... 21
3.4. Output Subsystem ......................................................................................................................................... 22
3.5. Extraction Engine Details ............................................................................................................................... 22
3.5.1. Getting Font Size Information ........................................................................................................................ 24
3.5.2. Searching Online Literature Databases ......................................................................................................... 25
V
3.5.2.1. Rule-based extraction from online literature databases ............................................................................... 29
3.5.3. HMM Model in this study .............................................................................................................................. 32
3.5.3.1. Word Features .............................................................................................................................................. 33
3.5.3.2. Smoothing method in this study .................................................................................................................... 33
3.5.3.3. Probability Calculation ................................................................................................................................. 34
3.5.3.4. Training HMM Model ................................................................................................................................... 35
3.5.3.5. Training Dataset ........................................................................................................................................... 36
3.5.3.6. Execution Example ........................................................................................................................................ 37
4. Experiment ........................................................................................................................................... 43
4.1.1. Experiment Dataset ...................................................................................................................................... 44
4.2. Performance Measurement ........................................................................................................................... 44
4.2.1. Comparison of different HMM state transition models ................................................................................. 45
4.3. Benchmark test ............................................................................................................................................. 47
4.3.1. Benchmark test 1 – HMM Result on CORA Header Dataset .......................................................................... 47
4.3.2. Benchmark test 2 – HMM Result on CORA References Dataset .................................................................... 48
4.3.3. Benchmark test 3 – HMM Author Extraction on Experiment Dataset ........................................................... 49
4.3.4. Benchmark test 4 – Font based HMM Title Extraction on Experiment Dataset ............................................. 49
4.3.5. Benchmark test 5 – Page 1 HMM Title Extraction on Experiment Dataset ................................................... 51
4.4. Entire System on Bibliography Extraction ...................................................................................................... 53
4.4.1. Performance for Overall System on Bibliography Extraction ......................................................................... 53
4.4.2. This study’s system versus Zotero .................................................................................................................. 55
5. Conclusion ............................................................................................................................................ 57
6. References ............................................................................................................................................ 58
Appendix A – List of scientific research used in real world extraction dataset .................................................... 61

1. Andrew, G. and Gao, J. (2007) Scalable training of L1-regularized log-linear models, International Conference on Machine Learning, 2007.
2. Bikel, D.M., Miller, S., Schwartz, R., and Weischedel, R. (1997) Nymble: a High-Performance Learning Name-finder, Proceedings of the Fifth Conference on Applied Natural Language Processing, 194-201.
3. Borkar, V., Deshmukh, K., and Sarawagi, S. (2001) Automatic segmentation of text into structured records, Proceedings of the 2001 ACM SIGMOD international conference on Management of data.
4. Center for History and New Media (2009) Zotero. http://www.zotero.org/, George Mason University.
5. Chen, S.F. (1998) An Empirical Study of Smoothing Techniques for Language Modeling, Technical Report, Computer Science Group, Harvard University, TR-10-98.
6. Covington, M.A. (2003) ET: an Efficient Tokenizer in ISO Prolog, Technical report, The University of Georgia, February.
7. Cowie, J. and Lehnert, W. (1996) Information extraction. Communications of the ACM, 39, 80-91.
8. Dugad, R. and Desai, U.B. (1996) A Tutorial on Hidden Markov Models. Technical Report No.: SPANN-96.1, May.
9. Feldman, R. and Sanger, J. (2007) The Text Mining Handbook, Cambridge University Press.
10. Freitag, D. (1998). Information Extraction from HTML - Application of a General Machine Learning Approach. Proceedings of the Fifteenth National Conference on Artificial Intelligence, 517-523.
11. Freitag, D. and Mccallum, A.K. (1999) Information Extraction with HMMs and Shrinkage, Proceedings of the AAAI-99 Workshop on Machine Learning for Information Extraction, 31-36.
12. Gale, W.A. and Church, K.W. (1994) What's wrong with adding one? In N. Oostdijk and P. de Haan, editors, Corpus-Based Research into Language. Rodolpi, Amsterdam.
13. Hetzner, E. (2004) A Simple Method for Citation Metadata Extraction using Hidden Markov Models, JCDL '08 Proceedings of the 8th ACM/IEEE-CS joint conference on Digital libraries
14. Jeffreys, H. (1948) Theory of Probability. Clarendon Press, Oxford, second edition.
15. Johnson, W.E. (1932) Probability: deductive and inductive problems. Mind, 41, 421-423.
16. Kohavi, R. and Provost, F. (1998) Glossary of Terms. Editorial for the Special Issue on Applications of Machine Learning and the Knowledge Discovery Process, 30.
17. Kupiec, J., Pedersen, J., and Chen, F. (1995) A Trainable Document Summarizer. Proceedings of the 18th annual international ACM SIGIR conference on Research and development in information retrieval.
18. Lidstone, G.J. (1920) Note on the general case of the Bayes-Laplace formula for inductive or a posteriori probabilities. Transactions of the Faculty of Actuaries, 8, 182-192.
19. Moens, M. (2006) Information Extraction: Algorithms and Prospects in a Retrieval Context, Springer.
20. Porter, M.F. (1980) An Algorithm for Suffix Stripping. Program, 14(3): 130–137
21. Qi, Y., Kiksa, P., Collobert, R., Sadamasa, K., Kavukcuoglu, K., and Weston, J. (2009) Semi-supervised sequence labeling with self-learned features. International Conference on Data Mining, 428-437.
22. Rabiner, L.R. (1989), A Tutorial on Hidden Markov Model and Selected Applications in Speech Recognition, Proceedings of the IEEE, 257-286.
23. Ram, A. and Moorman, K. (1999) Understanding language understanding: computational models of reading, MIT Press, Cambridge, MA.
24. Ramage, D. (2007) Hidden Markov Model Fundamentals, Machine Learning, Course Materials, Section notes 5.
25. Schmid, H. (2008), Tokenizing, In Anke Lüdeling and Merja Kytö, editors: Corpus Linguistics, An International Handbook. Mouton de Gruyter, Berlin.
26. Seymore, K., Mccallum, A., and Rosenfeld, R. (1999) Learning Hidden Markov Model Structure for Information Extraction. AAAI 99 Workshop on Machine Learning for Information Extraction
27. Shen, D., Sun, J., Yang, Q., and Chen, Z. (2007) Document summarization using conditional random fields. International Joint Conference on Artificial Intelligence, 2007, 2862-2867.
28. Turmo, J., Ageno, A., and Català, N. (2006) Adaptive information extraction, ACM Computing Surveys, 38, 10-47.
29. U.S. Government Printing Office (2008), U.S. Government Printing Office Style Manual, 30.
30. Viterbi, A.J. (2002)x Error bounds for convolution codes and asymptotically optimum decoding algorithm. IEEE Transactions on Information Theory 13 (2): 260–269.
31. Warakagoda, N.D. (1996) A hybrid ANN-HMM asr system with nn based adaptive preprocessing, Master's thesis, Institutt for Teleteknikk, Transmisjonsteknikk.
32. Yang, H., Onda, N., Kashimura, M., and Ozawa, S. (1999) Extraction of bibliography information based on image of book cover, Image Analysis and Processing, 1999.
33. Yin, P., Zhang, M., Deng, Z.H., and Yang, D.Q. (2005) Metadata Extraction from Bibliographies Using Bigram HMM, Lecture Notes in Computer Science, 2005, Volume 3334, Digital Libraries: International Collaboration and Cross-Fertilization, Pages 1-14
34. Zhang, M., Yang, D., Deng, Z.H., Feng, Y., Wang, W., Zhao, P., Wu, S., Wang, S., and Tang, S.W. (2004) PKUSpace: A Collaborative Platform for Scientific Researching, Advances in Web-Based Learning – ICWL 2004, Lecture Notes in Computer Science, Volume 3143/2004, 245-260.
35. 蕭文峰, 何偉豪, and 李姵璇. (2007), PDF書目資料擷取系統之建置, Proceedings of International Conference on Information Management.