醫學生要正確判斷病患的疾病是相當不容易的，必須要有充分的知識以及大量的實務練習才有可能做到。現今有越來越多醫學系課程採用問題導向式學習 (Problem-Based Learning, PBL) 的教學方式來訓練學生，希望利用這種情境式的教學方式讓學生把課堂上的內容與現實狀況連結起來。但是使用PBL的教學方式會耗費很多的人力和成本，為了減輕醫學院的負擔，在先前的研究中有開發一套「電腦化虛擬病人」（Computerized Virtual Patient, CVP），希望能運用此系統來提高學習的品質，減輕小組老師的負擔，並降低問題導向式學習的教學成本。在電腦化虛擬病人系統中，學生透過病史詢問模組提出問題詢問病患的病史資料，由系統回答出教案中相對應問題的答案。但是這套系統在虛擬病人的病史詢答上表現的不太好，找尋相對應問題的準確率並不是很高。
本研究提出一套方法來改善病史詢答系統，採用問題檢索的架構來做病史詢答。透過三大步驟：「問題前處理」、「問題分類」、「問題相似度比對」，處理輸入問題，希望能提高比對問題的準確率，並且可以降低系統出現參考問題的機率，讓學生不會只是依靠提示來猜測出最後的結果。「問題前處理」包含中文斷詞、長詞合併、同義詞取代、去除停詞以及語意類別取代等五個步驟。「問題分類」則是把經過前處理後的問句與類別句型樣式(pattern)做比對，來找出問題應該要屬於的類別。「問題相似度比對」是把分好類別的問題與教案中同類別的問題進行相似度比對，來找出與輸入問題相同意思的教案問題。在計算問題間的相似度時，除了使用向量空間模型餘弦相似性（cosine similarity）外，另外也考量語意類別詞的相似度分數，希望能夠提高問題比對的準確率。
實驗結果顯示當問題相似度閥值設為0.7時，系統正確回答問題的準確率有84.65%，證實我們採用的方法確實可以改善系統的效能，降低提示出現的情況，讓學生在使用此系統時能多加練習自己的邏輯思考推理能力。

There are more and more courses in medicine using problem-based learning (PBL) method to teach students. Students can strengthen their medical knowledge by practicing diagnosis through a designed problem. A PBL course needs many tutors to guide and to monitor the learning process of groups of students. In order to reduce the workload of tutors, Computerized Virtual Patient (CVP) system has been proposed to support the learning process. In CVP system, students gather medical history information of patients by asking the patient history inquery subsystem. In previous study, the performance of patient history inquery subsystem was not very good.
In this study, we try to enhance the patient history inquery subsystem. We adopt the architecture of question retrieval to find the answer of the student’s questions. Three steps are taken while answering a student’s question: question preprocessing, question classification, and question similarity calculation. Question preprocessing contains five steps: Chinese word segmentation, long word identification, synonym normalization, removing stop words, and replacement of semantic class. Question classification module uses question patterns to find the class a question belongs to. After classification, the similarities between the student’s question and the predefined questions in PBL problem with the same class as the student’s question are calculated. In addition to the cosine-similarity of vector-space-model, we also consider the overlapping words with the same semantic class in two questions while calculating question similarity.
Experimental results showed that the accuracy of finding correct question comes to 84.65% when the similarity threshold was set to 0.7. The high accuracy can reduce the chance of providing candidate questions to students. And students can develop their ability of logical reasoning and diagnosis skill more effectively.

摘要 I
Abstract II
目錄 III
圖目錄 V
表目錄 VI
一、 簡介 1
1.1 背景 1
1.2 研究動機與目的 3
二、 文獻探討 5
2.1電腦化虛擬病人 5
2.2問題分類與特徵選取 6
2.3相似度計算 7
三、 研究方法 10
3.1 問題前處理 12
3.1.1 問句斷詞 13
3.1.2 長詞合併 13
3.1.3 同義詞取代 13
3.1.4 去除停詞 14
3.1.5 語意類別取代 15
3.2 問題分類方法 16
3.2.1 定義問題類別 17
3.2.2 找出類別句型樣式 18
3.2.3 利用類別句型樣式進行問題分類 19
3.3 問題比對 19
3.3.1 相似分數計算 19
3.3.2 語意類別分數 21
四、 實驗評估 23
4.1 實驗資料 23
4.2 評估準則 25
4.3 基本方法：無前處理、無問題分類 25
4.4 加入問題前處理 26
4.5 加入問題前處理和語意類別分數 26
4.6 問題分類效能 28
4.7 加上問題分類後之問題檢索效能 29
4.8 判定是否是相同問題之閥值 30
4.9 教案測試 32
五、 結論與未來展望 34
參考文獻 36

[1]王淳莆, 王書妤, 楊凱博, 電腦化虛擬病人, 慈濟大學醫學資訊學系醫學資訊專題報告, 2009.
[2]吴乃龙, "最大熵方法," IEEE 北京分会第一届学术年会论文集, 1987.
[3]李茹, 宋小香, 王文晶, "基于汉语框架网的中文问题分类," 计算机工程与应用, Vol. 31, pp. 111-114, 2009.
[4]高欣群,健康資訊網站之中文醫療問題自動分類-以西醫為例, 慈濟大學醫學資訊研究所碩士論文, 2008.
[5]許麗齡, "問題導向學習於護理教育上之應用," 護理雜誌, Vol. 48, pp. 31-36, 2001.
[6]陸希平, 王本榮, 陳家玉, "問題導向學習," 醫學教育, Vol. 10, pp. 89-97, 2006.
[7]賴育昇, 李坤霖, 吳宗憲, "網際網路 FAQ 檢索中意圖萃取與語意比對之研究," in Proceedings of Research on Computational Linguistics Conference XIII, pp. 135-155, 2000.
[8]M. A. Ainsworth, L. P. Rogers, J. F. Markus, N. K. Dorsey, T. A. Blackwell, and E. R. Petrusa, "Standardized Patient Encounters: A Method for Teaching and Evaluation," The Journal of the American Medical Association, Vol. 266, pp. 1390-1396, 1991.
[9]O. Bodenreider, "The Unified Medical Language System (UMLS): Integrating Biomedical Terminology," Nucleic Acids Research, Vol. 32, pp. D267-D270, 2004.
[10]J. S. Brown, A. Collins, and P. Duguid, "Situated Cognition and the Culture of Learning," Educational Researcher, Vol. 18, pp. 32-42, 1989.
[11]R. Burke, K. Hammond, V. Kulyukin, S. Lytinen, N. Tomuro, and S. Schoenberg, "Natural Language Processing in the FAQ Finder System: Results and Prospects," in Working Notes from AAAI Spring Symposium on NLP on the WWW, pp. 17-26, 1997.
[12]D. Cai, J. Sun, G. Zhang, D. Lv, Y. Dong, Y. Song, and C. Yu, "HowNet Based Chinese Question Classification," in Proceedings of the 20th Pacific Asia Conference on Language, Information and Computation, pp. 366-369, 2006.
[13]Z. Dong and Q. Dong, HowNet and the Computation of Meaning, World Scientific, 2006.
[14]S. Dumais, J. Platt, D. Heckerman, and M. Sahami, "Inductive Learning Algorithms and Representations for Text Categorization," in Proceedings of the Seventh International Conference on Information and Knowledge Management, pp.148-155, 1998.
[15]P. Indyk and R. Motwani, "Approximate Nearest Neighbors: Towards Removing the Curse of Dimensionality," in Proceedings of the Thirtieth Annual ACM Symposium on Theory of Computing, pp. 604-613, 1998.
[16]J. M. Keller, M. R. Gray, and J. A. Givens, "A Fuzzy K-Nearest Neighbor Algorithm," IEEE Transactions on Systems, Man and Cybernetics, Vol. SMC-15, pp. 580-585, 1985.
[17]D. M. Magerman, "Statistical Decision-Tree Models for Parsing," in Proceedings of the 33rd Annual Meeting of the Association for Computational Linguistics, pp. 276-283, 1995.
[18]G. A. Miller, "WordNet: A Lexical Database for English," Communications of the ACM, Vol. 38, pp. 39-41, 1995.
[19]M. Mishra, V. K. Mishra, and H.R. Sharma, "Question Classification using Semantic, Syntactic and Lexical Features," International Journal of Web & Semantic Technology, Vol. 4, pp. 39-47, 2013.
[20]D. Mladenic and M. Grobelnik, "Feature Selection for Unbalanced Class Distribution and Naive Bayes," in Proceedings of the 16th International Conference on Machine Learning, pp. 258-267, 1999.
[21]S. Niwattanakul, J. Singthongchai, E. Naenudron, and S. Wanapu, "Using of Jaccard Coefficient for Keywords Similarity," in Proceedings of the International MultiConference of Engineers and Computer Scientists 2013, pp. 380-384, 2013.
[22]E. Riloff and M. Thelen, "A Rule-based Question Answering System for Reading Comprehension Tests," in Proceedings of the 2000 ANLP/NAACL Workshop on Reading Comprehension Tests as Evaluation for Computer-Based Language Understanding Sytems, Vol. 6, pp. 13-19, 2000.
[23]S. E. Robertson, S. Walker, S. Jones, M. M. Hancock-Beaulieu, and M. Gatford, "Okapi at TREC-3," Overview of the Third Text REtrieval Conference (TREC-3), pp. 109-126, 1995.
[24]G. Salton and C. Buckley, "Term-Weighting Approaches in Automatic Text Retrieval," Information Processing & Management, Vol. 24, pp. 513-523, 1988.
[25]G. Salton, A. Wong, and C.S. Yang, "A Vector Space Model for Automatic Indexing," Communications of the ACM, Vol. 18, pp. 613-620, 1975.
[26]J.G. Sun, D.F. Cai, D.X. Lv, and Y.J. Dong, "HowNet Based Chinese Question Automatic Classification ," Journal of Chinese Information Processing, Vol. 21, pp. 90-95, 2007.
[27]C. H. Wang, G. G. Yang, R. L. Liu, W. C. Lin, and C. J. Lin, "Developing Computerized Virtual Patients in PBL for Medical Students at Tzu-Chi University," in Proceedings of the 1st Asia-Pacific Joint PBL Conference, pp. 93, 2010.
[28]D. F. Wood, "Problem Based Learning," British Medical Journal, Vol. 326, pp. 328-330, 2003.
[29]D. Zhang and W. S. Lee, "Question Classification using Support Vector Machines," in Proceedings of the 26th Annual International ACM SIGIR Conference on Research and Development in Informaion Retrieval, pp. 26-32, 2003.
[30]Z. Zhou, S. Shi, Y. Han, and X. Lv, "Similarity Computing Between Simple Questions Based on HowNet and Syntactic Analysis," in Proceedings of the. International Conference on Asian Language Processing, pp. 28-33, 2009.