臺灣博碩士論文加值系統

本研究以資訊擷取及自然語言處理的技術，將非結構化自由型態消化道內視鏡報告，以消化道內視鏡基礎標準術語(MST2)為本體，經過分析整理成「可延伸性標示語言」(Extensible Markup Language，簡稱XML)的消化道內視鏡報告，提供查詢，跨平台交換及醫學統計。未來也可與消化道內視鏡影像儲存系統結合，合成完整消化道內視鏡電子資料庫系統，提供醫學照顧與治療決策更精確的資訊。
研究方法：首先，評估國內消化道內視鏡檢查報告的結構性符合MST2架構的程度，依據MST2架構將消化道內視鏡檢查報告依檢查的部位、檢查的發現及診斷分三個部分，每一個部份有三個評估子項，每一子項依據與MST2符合程度給分，由完全符合給與3分至完全不符合給與0分，由一位消化系專科醫師對每篇文章評分，每篇報告最高總分共27分。其次，進行名詞剖析，以MST2同義字引擎將已剖析名詞進行同義字轉換成MST2標準名詞，處理文章切割則將文章依據四個文章中的關鍵詞 Esophagus、Stomach、Duodenum及Impression將文章分為以此四個字為首的四個區段。計算每一個標準名詞在各區段文章中的序位。將轉換後擷取出的標準名詞以XML進行文字索引及標示，最後形成以MST2為索引典的結構式自動分類內視鏡報告。預期經過處理後的消化道內視鏡報告其文件分類可以更為有效。評估方法將以公元2001年由某市立醫院5位消化系專科醫師紀錄的1266篇消化道內視鏡報告為訓練資料庫，以2005年某市立醫學中心1699篇檢查報告為測試資料庫，其中由六位消化系專科醫師撰寫的消化道內視鏡報告489篇及上消化道治療性內視鏡檢查報告30篇。以“回現率”與“精確率”二因素來衡量評估檢索效能。程式語言使用JAVA。轉換後檢查報告以XML格式儲存。
結果：選取1235篇訓練組資料庫進行MST2結構式相符評分分析，發現只有33 篇(2.7%)報告檢查內容能完全符合標準名詞規範，但是大部分(60%)的內視鏡報告9個子項中仍有4到6項符合MST2的要求。將訓練資料庫1266篇文字進行同義字分析後，設計十種MST2同義字轉換形式，發現轉換後可增加訓練資料庫標準名詞的擷取數達2.56倍以上。訓練資料庫與測試資料庫皆能以超過95%的精確率與回現率來將文章正確的切割成為以esophagus、stomach、duodenum與impression為段首的四個區段。若以stomach、ulcer及body三個標準名詞來評估文章檢索功能，將訓練資料庫調教至精確率為76.9%、回現率100%時，在測試資料庫精確率為68.75%、回現率100%。若能以”esophagus/varices/grade II/lower third”固定序位的次序模式擷取，文件檢索效能在訓練資料庫會由精確率:80.7%、回現率: 74.2%變為精確率:93.6%(44/47)、回現率: 70.9%(44/62)。
結果與討論：對非使用英文為母語的醫師，無法很精確使用英文準名詞來描述消化道內視鏡檢查結果，只有60%的檢查報告內容能部份符合MST2標準名詞規範，很少(2.7%)能完全符合，因此這類報告在資訊擷取或文章分析時，將無法得到正確結果。但是消化道內視鏡檢查內容皆具有一固定格式，因此可依據關鍵詞來進行文章分割，有超過95%的精確率與回現率。不過當有不同類型報告混合進行分析時，如將超音波或大腸鏡檢查報告一起分析時，精確率會下降至60%以下。在醫學報告中進行名詞檢索，若先經過同義字轉換成標準名詞後，有非常高的回現率，但是精確率只有70%，原因在於醫學專有名詞較統一且經過同義字轉換，容易擷取，但是檢索的名詞間是否在文章中有相關，則需經過自然語言處理來協助提高精確率。MST2標準名詞在報告內容中會有固定的出現序位，將序位模式列入分析時，可增進檢索精確率。同樣發現在一些檢查為正常的報告，醫師會以固定報告範本輸入，如檢索ulcer時，可發現ulcer出現於文章中第18個位置時，內容為no ulcer的可能有95%。但是當擷取名詞時，有多個從屬名詞出現時，如bleeding、erosion與ulcer都被檢索出時，需藉助三個名詞彼此間在文章中序位的向量分析，才能正確分類。自由型態非結構式消化道內視鏡檢查報告在自然語言處理後，是可以轉換成標準名詞結構式報告，且可進行檢索與分類。如此，可以達到更有效的分類，本研究發現轉換後的消化道內視鏡檢查報告分類，食道區的主要病變為Erosion及Ulcer且大都是單一病灶出現；在胃部主要病灶為多發的erosion，或是單一的superficial ulcer且沒有Bleeding；在12指腸以單一的cratered ulcer為主。不過分類仍會因為錯別字，自創簡寫及負面敘述，造成錯誤，需有更強大的字典庫及同義字庫與人工智慧來協助分辨，文字序位向量及文章模式或許可作為人工智慧分析的條件。
結論：在分析不是以英文為母語的文章撰寫者而言，不論在文字、文法及描述上有相當大的認知差距，再加上有自創縮寫及包括格式及打字錯誤的書寫錯誤，造成自然語言處理的困難，若能用標準名詞及架構、導入概念空間(Concept Space)及知識主體(Domain Knowledge)的協助，資訊擷取及文件將可更精確。本研究嘗試用世界消化內視鏡醫學會頒布的消化道內視鏡基礎標準術語第二版(MST 2)名詞作為同義字轉換及結構式資料庫轉換依據，以可延伸性標示語言〈XML〉作為轉換後標準語言格式，提供一種可行方式將非結構化自由型態消化道內視鏡報告以自然語言處理轉換成標準名詞結構且自動分類的消化道內視鏡報告，提高資訊擷取與文件分析的效能。

In this research we used information retrieval (IR) technique and natural language processing (NLP) to transform nonstructural free style gastroendoscopic reports into structured formatted with the extensive markup language (XML) based on the minimal standard terminology version 2 (MST2) ontology. These structured reports can be used for classification, query, sharing and medical decision making. These reports and their counterparts, the endoscopy images, can be integrated as an intelligent endoscopic report system to support more accurate and precise information for decision making on care and treatment.
Study design: The terms proposed in for describing location of findings, terms, and diagnosis were used to match the ones in the gastroendoscopy reports. The MST2 terms are classified into three categories. The match score was evaluated based on the fitness of the terms described in each subcategory with MST2 from complete match scored 3 to no match scored 0.The total scores are 27. Each report was processed by the techniques including synonym transformation, segmentation, ordering, indexing, marking up and classification. Synonym transformation technique is used to map the local terms to the corresponding MST2 terms. A report is segmented into four paragraphs by four leading words: Esophagus, Stomach, Duodenum and Impression. We marked up the standard MST2 terms and calculated each standard MST2 term’s position in each segmented paragraph. Finally, we transformed the reports into structured gastroendoscopic reports. We evaluated the power of information retrieval and text analysis by using the precision rate and recall rate. We selected 1266 gastroendoscopy reports recorded by five gastroenterologists in 2001 in a municipal hospital as the training set and other 1699 medical examination reports, in which 489 gastroendoscopy reports and 30 therapeutic endoscopy reports recorded by 6 gastroenterologists, from a municipal medical center as the testing set. The programs were written by JAVA language and the reports were stored as XML files.
Results: There were only 33 (2.7%) reports that matched the MST2’s standard terms and ontology in 1235 reports from the training set, but 60% of the reports could score from 4 to 6 in each category. We analyzed 1266 reports from the training set and developed ten synonym templates based on MST2 standard terms. Using the synonym templates for extraction of MST2 terms from the reports, we could increase the number of terms up to 2.56 folds compare to those without the templates. There were higher than 95% precision rate and recall rate to segment those reports from both the training set and testing set into 4 segments. We evaluated information retrieval function with stomach, ulcer and body. The training set tuned the precision rate as 76.9% and the recall rate as 100%. The results of the testing set were precision rate 68.75% and recall rate 100%. We set the extraction model in fixed orders as “esophagus/varices/grade ii/lower third”. We modified retrieval function from precision rate 80%, recall rate 74.2% to precision rate 93.6% and recall rate 70.9% .
Discussion: It is discovered that nonnative English doctors could not precisely use English standard terms to prescribe findings in the gastroendoscopy examinations. Hardly to extract accurate information from these reports. But, there were still some fixed formats in gastroendoscopic reports may segment it into designed paragraphs with high (>95%) precision rate and recall rate. If the data sets mixed with various reports from colonoscopy or sonography examinations, the precision rate will decrease lower than 60%. We could retrieval terms with high recall rate after synonym transformation to be standard terms but lower precision rate (70%) in medical examination reports. The reasons were that the terms we retrieved may not being relevant. We need natural language processes to parse the terms to get higher precision rate. There were fixed position orders in some gastroendoscopic reports. If we input some terms’ orders then we could improve our retrieval precision rate. We found that physicians recorded normal reports with fixed template. There were 144 reports recorded ulcer in the position 18 of stomach paragraph and 95% reports associated with no ulcer. When we retrieved more than one terms with relevant such as bleeding, ulcer and erosion, we need vector analysis with position among them and to get more accurate precision rate and classification rate.. We can process free style nonstructural gastroendoscopic reports by natural language to be a structured standard reports. Also errors may cause by erratum, personal abbreviations and negative descriptions. We need more powerful medical dictionary and synonyms map to parse the terms written by nonnative English doctors. Word’s position vectors and report’s models could be the factors for artificial intelligence. We analyzed automatic classified structured reports according to NLP algorithm. The main lesions in Esophagus segments were a single erosion or ulcer. The main lesions in Stomach segments were multiple erosions or a single superficial ulcer without bleeding. The main lesion in Duodenum segments was a single cratered ulcer.
Conclusion: It was difficult to analyze the medical reports recorded by those nonnative English doctors due to the large discrepancy in terms, grammar and prescriptions, There were personal abbreviations and wrong spellings including formatted or typed errors that made natural language processing more difficult. What if we could use standard terminology concept space and domain knowledge that we could extract information more accurately. This research selected World Organization for Digestive Endoscopy (OMED) published Minimal Standard Terminology Version 2 for synonyms and structured transformation’s templates and transformed the gastroendoscopic reports to XML reports. These transformations could improve information retrieval and text analysis accuracy. The standard terms synonym transformation and markup with XML reports could be a useful nature language process to analyze the nonstructural free style gastroendoscopic reports.

目錄

頁數
標題 i
審定書 ii
上網授權書 iii
國科會授權書 iv
誌謝 vi
目錄 vii-viii
表目次 ix-x
圖目次 xi- xii
中文摘要 xiii- xv
英文摘要 xvi-xvii
第一章 緒論
1.1 研究背景 1
1.2 研究動機 1-11
1.3 研究目的 11-12
第二章 文獻查證
2.1 醫學文件整理與分類 13-17
2.2 醫學文件辨識系統 18-22
2.3 消化道內視鏡報告資訊擷取 22-26
第三章 研究材料與方法
3.1 資料庫 26-28
3.2 搜尋引擎設計 28-39
3.3 非結構式消化道內視鏡報告分析 39-40
3.4 自然語言文字擷取模式 40-43
3.5 結果分析方法 43
第四章 分析與結果
4.1 非結構式消化道內視鏡報告分析結果 44-46
4.2 同義字轉換分析 47-50
4.3 文章分段分析結果 51-53
4.4 將非結構性消化道內視鏡報告轉成標準名詞結構式資料庫 54-58
4.5 資訊擷取功能評估 59-61
4.6 以序位作為分析條件的資訊擷取功能評估 61-62
4.7 自動分類結果分析 62-66
第五章 討論
5.1 影響醫師在撰寫自由文件格式的英文消化道內視鏡報告所適用的辭彙的因素 67
5.2 自然語言處理技術的難題 68-71
5.3 產生可延伸性標示語言消化道內視鏡報告的問題 71
5.4 以消化道內視鏡基礎標準術語第二版(MST 2)名詞進行同義字轉換的問題 71-72
5.5 如何分辨負面敘述 72
5.6 以不同區段相同或相關關鍵詞做擷取，可增加擷取率，也發現因為不用標準名詞造成前後文不一致現象 72-73
5.7 如配合人工智慧來訓練才可能會有較佳成果 73-74
第六章
結論與建議 74-75
參考資料
中文文獻 76-77
英文文獻 78-82
電子資料 82-84

中文文獻
1.黃雲龍.(1998):中文全文資訊檢索研究架構與重要議題探討;大學圖書館 University Library Quarterly 第二卷第三期，民87。
11.謝清俊(1994)，「語文工作與資訊發展－－從電子文件的發展談對語文研究的期盼」，當前語文問題學術研討會，(台北：行政院國家科學委員會，國立台灣大學中國文學系)，附錄，民83。
14.黃雲龍 (1997)，「中文全文文件群集索引理論研究--向量空間模型(Vector-Space Model)的建構」，(國立台灣大學商學研究所博士論文，民國86年)，頁7，民86。
24.簡立峰，(1996)「尋易系統(Csmart)與中文智慧型資訊檢索」，21世紀資訊科學與技術的展望國際學術研討會論文集，(台北市：世界新聞傳播學院圖書資訊學系，國家圖書館)，頁299-314，民85。
25.張俊盛，柯淑津，陳惠群，(1998)「跨語言資訊檢索的網路資源探勘」，21世紀資訊科學與技術的展望學術研討會論文集，(桃園縣：世界新聞傳播學院圖書資訊學系)，頁153-166，民87。
26.曾元顯，(1996)「多媒體資訊檢索技術之探討」，21世紀資訊科學與技術的展望國際學術研討會論文集，(台北市：世界新聞傳播學院圖書資訊學系，國家圖書館)，頁281-298，民85。
27.Chen, Hsin-His, (1996) "Cross-Language Information Retrieval", 電子詞典、機器翻譯與資訊擷取研討會，(台北市：中華民國計算語言學會) ，民86。
28.陳光華，(1997)「電子文件主題之自動辨識」，中國圖書館學會會報五十九期，頁43-58，民86。
29.曾元顯，(1997)「關鍵詞自動擷取技術與相關回饋」，中國圖書館學會會報五十九期，頁59-64，民86。
30.卜小蝶，(1996)「網路資訊過濾技術與個人化資訊服務」，21世紀資訊科學與技術的展望國際學術研討會論文集，(台北市：世界新聞傳播學院圖書資訊學系，國家圖書館)，頁339-350，民85。
31.卜小蝶，(1997)「提供個人化服務的線上公用目錄檢索系統初探」，中國圖書館學會會報五十九期，頁127-134，民86。
32.陳光華(1998)，「新資訊時代的啟發性資訊服務」，21世紀資訊科學與技術的展望學術研討會論文集，(桃園縣：世界新聞傳播學院圖書資訊學系，頁195-208，民87)。
33.林頌堅，(1998)「自動化文件分類在資訊服務上的應用」，21世紀資訊科學與技術的展望學術研討會論文集，(桃園縣：世界新聞傳播學院圖書資訊學系)，頁255-275，民87。
35.朱學鋒，俞士汶，羅振聲，袁毓林‧；自然語言處理與語言知識庫*《計算機時代的漢語漢字研究》，清華大學出版社(1996)。
37.蘇諼，(1997) 臺灣地區臨床醫師的資訊需求與資訊尋求模式研究(台北市：國科會)，頁1，民86年.
39.張嘉彬 (1998). 索引典及其於資訊檢索上之探討.書苑季刊36期 第46-59頁，民87。
43.美國資訊科學學會臺北分會，(1994)＜索引典理論與實務＞（臺北市：美國資訊科學學會臺北分會），頁8，民83.10。
44.蔡明月，(1991)＜線上資訊檢索：理論與實務＞（臺北市：學生），頁177，民80。
45.黃惠株，(1996)「淺談索引典」，＜佛教圖書館館訊＞第五期，頁2，民85.3。
46.陳攸華,(1995)＜圖書資訊學研究＞（臺北市：文華）頁34-35，民84。
47.黃惠株，(1996)「淺談索引典」，＜佛教圖書館館訊＞第五期，頁2-3民85.3。
48.黃惠株，(1996)「淺談索引典」，＜佛教圖書館館訊＞第五期，頁3-5，民85.3。
49.美國資訊科學學會臺北分會(1994)，＜索引典理論與實務＞（臺北市：美國資訊科學學會臺北分會），頁47-48，民83.10。
50.張嘉彬. ，(1998) 索引典及其於資訊檢索上之探討.書苑季刊36期 第46-59頁，民87。
56.阮仲洲等/(Maratka, Zdenek), (1999) 消化系內視鏡 : 用語定義與診斷標準。藝軒(415.41 7152 88) ，民88。
65.陳光華.(1996)"資訊檢索查詢之自然語言處理"，中國圖書館學會會報，第57 期，頁 141–153，民 85.12。
68.陳光華，陳信希。(1996)「文件內容之分析--語料庫為本的模型」。台灣大學圖書館學刊第十一期，頁97-114，民85.12。
70.蔡明月，(1991) ＜線上資訊檢索：理論與實務＞（臺北市：學生），頁168-169，民80。
88.黃慕萱，(1996) ＜資訊檢索＞（臺北市：學生），頁4，民85
89.美國資訊科學學會臺北分會，(1994) ＜索引典理論與實務＞（臺北市：美國資訊科學學會臺北分會），頁26-29，民83.10。

英文文獻
2.Salton G. (1989) Automatic Text Processing: The Transformation, Analysis, and Retrieval of Information by Computer, Addison-Wesleym.
3.Korman L, Crespi M, Delvaux M (2000). Minimal Standard Terminology in Digestive Endoscopy. Endoscopy 32(2): 162-88.
4.Hull R, King R (1987). Semantic Database Modeling: Survey, Application, and Research Issues. ACM Computing Surveys. 19(3): 201-60.
5.Saffady W (1989). Text Storage and Retrieval Systems: A Technology Survey and Product Directory (Meckler Corporation).
6.Salton G. (1991). The SMART document retrieval project, Proc. Fourteenth International ACM/SIGIR Conference on Research and Development in Information Retrieval, 56-8.
7.Air Force Systems Command. Research and Technology Division. Rome Air Development Center. Intelligence and Information Processing Division. (1967) Handbook for Planning File Conversion,RADC-TR-67-168, NBS# 72287012. (Box 570, folder 11).
8.IBM World Trade Corp. (1971) STAIRS: Storage and Information Retrieval System, General Information Manual, IBM No. GH12-5707, IBM System/360 and System/370 (OS) Storage and Information Retrieval System, booklet GH12-5107-D.
9.IBM. (1972)(STAIRS) - General Information Manual, Second Edition, Stuttgart:。
10.Saffady W (1989). Text Storage and Retrieval Systems: A Technology Survey and Product Directory, Meckler Corporation.
12.Van Rijsbergen CJ (1983). "Information Retrieval: New Directions: Old Solutions," in Proceedings of The ACM SIGIR. 264.
13.Salton G. (1986). Another Look at Automatic Text-Retrieval Systems, Communications of the ACM, 9(7): 651.
15.Salton G. (1964). A Flexible Automatic System for the Organization, Storage, and Retrieval of Language Data (Smart), Scientific Report ISR-5, Section I, Harvard Computation Laboratory, Cambridge MA.
17.Salton G.. (1971). The SMART Retrieval System, Experiments in Automatic Document Processing (Prentice Hall, Inc., Englewood Cliffs, N. J.,).
18.Lewis DD (1992). An Evaluation of Phrasal and Clustering Representations on a Text Categorization Task, in Proceedings of The ACM SIGIR, 37-50.
19.Deerwester S, Dumals ST, Furnas G.W, Landauer TK, Karshman R (1990). Indexing by Latent Semantic Analysis, JASIS 41(6): 391-407.
20.Can F (1994). The Efficiency of Best-Match Cluster Searches, Information Processing & Management, 30(3): 343-61.
21.Wilkinson R, Hingston P (1991). The Cosine Measure in A Neural Network for Document Retrieval, in Proceedings of ACM-SIGIR, 202-10.
22.Lu X (1990). Document Retrieval: A Structural Approach, Information Processing & Management 26(2): 209-18.
23.Crouch CJ (1990). An Approach to The Automatic Construction of Global Thesauri, Information Processing & Management 26(5): 632.
36.Osheroff JA, et al (1991). Physicians’ information needs: analysis of questions posed during clinical teaching, Annals of Internal Medicine, 114(7): 576。
38.Stavri Z (1996). Medical problem attributes and information-seeking questions, Bulletin of the Medical Library Association, 84(3): 367-8。
40.Mooers CN (1959). Information Retrieval Selection Study. Part II: Seven System Models. Cambridge, MA: Zator Company.
41.Mooers CN (1959). Mooers’ Law; or why some retrieval systems are used and others are not. (Zator Technical Bulletin 136). Cambridge, MA: Zator Company; 1959. Editorial of same title, American Documentation, 11(3): i; July 1960; and reprinted in The Scientist, 11(2): 10, Mar 17, 1997.
42.Austin B. Mooers’ Law: In and Out of Context; the Journal of the American Society for Information Science and Technology (JASIST), 2001; 52(8).
51.Lindberg DA, Humphreys BL, McCray AT (1993). The Unified Medical Language System, Methods of Information in Medicine, 32(4): 281
52.Maratka, Z (1984). Terminology, definitions and diagnostic criteria in digestive endoscopy. Scan J Gastroenterol, 19(suppl 103): 1-74.
53.Maratka Z (1994) Terminology, Definitions and Diagnostic Criteria in Digestive Endoscopy, Third Edition. Bad Homburg: Normed Verlag.
54.Crespi M, Delvaux M, Schapiro M, Venables C, Zweibel F (1994). Minimal standards for a computerized endoscopic database. Am J Gastro. 89: S144-S153.
55.Maratka Z (1999). Endoscopic diagnosis in gastroenterology, 4th ed. Bad Homburg, Germany: Normed.
57.WHO-CIOMS Terminology Circular No. IND-1 October 1978.
58.International Statistical Classification of Diseases and related Health Problems (ICD-10).10. Revision, Vol. 1, WHO Geneva, 1992.
59.Sperberg-McQueen CM. XML and the future of digital libraries。The Journal of Academic Librarianship 24: 4 (1998 July): 314-7
61.DeBakey ME (1991). The National Library of Medicine: evolution of a premier information center, JAMA 266(9): 1251-8.
63.Hersh WR, Hickam D (1995). Information retrieval in medicine: The Sapphire experience. J Am Soc Inf Sci. 46(10): 743-7.
66.Aronow DB, Fangfang F, Croft WB (1999). Ad hoc classification of radiology reports. J Am Med Inform Assoc. 6(5): 393-411.
68.Walter P, Jonathan S, Thomas L (1999). Technologies for Extracting Full Value from the Electronic Patient Record; Proceedings of the 32nd Hawaii International Conference on System Sciences.
70.Kestenbaum, Peter (1995). Conceptually Understanding JAVA in ten Minutes without Getting Lost in Cyberspace, JAVA BOOK, SUN Microsystems, Inc.
71.Hahn U, Romacker M, Schulz S, (2000) medSynDiKATe: Design considerations for an ontology-based medical text understanding system. In: AMIA 2000 - Proc. Annual Symposium of the American Medical Informatics Association. Converging Information, Technology, and Health Care. Los Angeles, CA, November 4-8, 2000. Ed. by J.M. Overhage. Philadelphia/PA: Hanley & Belfus, 330-34.
72.Sager N, Lyman M, Nhan NT, TIck LJ (1995). Medical language processing: applications to patient data representation and automatic encoding. Meth Inform Med; 5; 34: 140-6.
73.Baud RH, Rassinoux AM, Wagner JC, Lovis C (1995). Representing clinical narratives using conceptual graphs. Meth Inform Med; 1/2: 176-86.
74.Friedman C, Alderson PO, Austin J, Cimino JJ, Johnson SB (1994). A general natural language text processor for clinical radiology. JAMIA; 1: 161-74.
75.Haug PJ, Ranum DL, Frederick PR (1990). Computerized extraction of coded findings from freetext radiologic report. Radiology, 174: 543-48.
76.Gundersen ML, Haug PJ, Pryor TA (1996). Development and evaluation of a computerized admission diagnoses encoding system. Computers and Biomedical Research, 29: 351-72.
77.Friedman C, Knirsch CA, Shagina L, Hripcsak G (1999). Automating a severity score guideline for community-acquired pneumonia employing medical language processing of discharge summaries. In Lorenzi N, ed. Proc AMIA Symp. Phil. Hanley & Belfus; 256-60.
78.Blanquet A and Zweigenbaum P (1999). A lexical method for assisted extraction and coding of ICD-10 diagnoses from free text patient discharge summaries. . In Lorenzi N, ed. Proc AMIA Symp. Phila. Hanley & Belfus; 1029.
79.Jain NL, Friedman C (1997). Identification of findings suspicious for breast cancer based on natural language processing of mammogram reports. In Masys, DR, ed. Proc AMIA Symp. Phila. Hanley & Belfus; 829-33.
80.Lussier Y, Shagina L, Friedman C (2000). Automated ICD-9 encoding using medical language processing: a feasibility study. Submitted to AMIA Fall Symp.
81.Friedman C, Starren J, Johnson SB (1995). Architectural requirements for a multipurpose natural language processor in the clinical environment, In: Gardner RM, ed. Proce of SCAMC 1995. Phil: Hanley & Belfus; 347-51.
85.Friedman C, Hripcsak G, DuMouchel W, Johnson SB, Clayton PD (1995). Natural language processing in an operational clinical information system. Nat Lang Eng; 1: 83-108.
86.Friedman C (1997). Towards a comprehensive medical language processing system: methods and issues. In Masys, DR. Proc AMIA Symp. Phil, Hanley & Belfus; 595-9.
87.Cleveland DB, Cleveland AD (1990). Introduction to Indexing and Abstracting (Englewood, Colo.: Libraries Unlimited), 95
90.Gilchrist A. The Thesaurus in Retrieval (1971), (London：ASLIB), 15-6。
91.Logan J, Klopfer K (2000). The use of a standardized terminology for comparison of free text and structured data entry. JAMIA Symposium Supplement.
92.Korman LY, Delvaux M, Bidgood D (1998). Structured reporting in gastrointestinal endoscopy: integration with DICOM and minimal standard terminology. International Journal of Medical Informatics. 48(1-3): 201-6.
93.Kuhn K, Swobodnik W, Johannes RS, Zemmler T, Stange EF, Ditschuneit H, Classen M (1991). The Quality of Gastroenterological Reports Based on Free Text Dictation - an Evaluation for Endoscopy and Ultrasonography. Endoscopy; 23: 262-4
94.M. Sackmann, et al (1998). A Scientific Relational Database Combined with a Report Generator for Endoscopy in Networks: Endonet, Endoscopy, 30(7): 610-6
95.Delvaux MM, Crespi M. Armengol-Miro JR, Hagenmuller F, Teuffel W. (1999) The GASTER project: building a computer network in digestive endoscopy: the experience of the European Society for Gastrointestinal Endoscopy. Gastrointestinal Endoscopy Application for Standards in Telecommunication, Education and Research. Journal of Clinical Gastroenterology. 29(2): 118-26.
96.Delvaux M, Crespi M, Armengol-Miro JR, et al (2000). Minimal Standard Terminology for digestive endoscopy: results of prospective testing and validation in the GASTER project. Endoscopy; 32: 345–55.
97.Kuhn K, Swobodnik W, Johannes RS, et al (1991). The quality of gastroenterological reports based on free text dictation: an evaluation in endoscopy and ultrasonography. Endoscopy 23(5): 262-4.
98.Kuhn K, Gaus W, Wechsler JG, et al (1992). Structured reporting of medical findings: evaluation of a system in gastroenterology. Methods of Information in Medicine, 31(4): 268-74.
99.Gouveia-Oliveira A, Raposo VD, Salgado NC, Almeida I, Nobre-Leitao C, de Melo FG (1991). Longitudinal comparative study on the influence of computers on reporting of clinical data. Endoscopy 23(6): 334-7.
100.Maratka Z (1989). Terminology, Definitions and Diagnostic Criteria in Digestive Endoscopy. Bad-Homburg: Normed-Verlag.
101.Korman L (1999). Standardization in Endoscopic Reporting. Journal of Clinical Gastroenterology 28(3): 217-23.
102.Logan J, Klopfer K (2000). The use of a standardized terminology for comparison of free text and structured data entry. JAMIA Symposium Supplement.
98.Fujita R. Endoscopic diagnosis and classification of esophageal varices in Japan, in Sivak MV’jr (ed): Sclerotherapy of Esophageal Varices. New York, Praeger, 1984, 35-42.
99.Cales P, Zabotto B, Meskens C, Caucanas JP, Vinel JP, Desmorat H, Fermanian J, Pascal JP (1990). Gastroesophageal endoscopic features in cirrhosis. Observer variability, interassociations, and relationship to hepatic dysfunction. Gastroenterology 98: 156-162.
100.Borrmann R, et al (1926). Handbuch der speziellen pathologischen Anatomic und Histologic Vol 4. Berlin: Springer, 865.
101.Pan S, et al (1981). An endoscopy study on the duodenal ulcer: An endoscopic classification of the duodenal ulcer and its implications. J Formosan Med Assoc 80: 815-29.
102.Fritz N, et al (2001). Compliance with terminology standards in reflux, ulcers, and gastritis: A study of 881 consecutive upper gastrointestinal endoscopy reports. Z Gastroenterol. 39(12): 1001-6.
103.Michele T, William T. H, Suresh S, Integration of a Standard Gastrointestinal Endoscopy Terminology in the UMLS Metathesaurus. Proceedings of the AMIA 2002 Annual Symposium, 801-5.
104. C, et al. Automated Encoding of Clinical Documents Based on Natural Language Processing. J Am Med Inform Assoc. 2004 Sep-Oct; 11(5):392-402. Epub 2004 Jun 7.
114.Logan J, Klopfer K (2000). The use of a standardized terminology for comparison of free text and structured data entry. JAMIA Symposium Supplement.

電子資料
16.Fox E (1983). some Considerations for Implementing the SMART Information Retrieval System Under UNIX. (Technical Report 83-560, Cornell University Department of Computer Science, Ithaca, New York.). Available at: http://cs-tr.cs.cornell.edu/. Accessed Nov. 26, 1996.
34.黃雲龍(1998)，中文全文資訊檢索研究架構與重要議題探討, A Study of Research Framework and Critical Issues concerning Mandarin Chinese Full Text Information Retrieval.大學圖書館 University Library Quarterly, 第二卷第三期 中華民國87年7月Vol.2, No.3. Available at: http://www.lib.ntu.edu.tw/pub/u nivj/uj2-3/uj2-3_1.htm. Accessed Jun. 25, 2005.
36.張俊盛，顯性知識管理的核心問題（下）。 Available at: http://www.gss.com.tw/eis/29/p55.htm Accessed Jan. 26, 2005.
40.元智大學，如何構建適當的檢索策略？Available at: http://www.yzu.edu.tw/library/4/7_1.htm Accessed Jul. 23, 2005.
60.陳葦真，XML對於圖書館自動化的應用. Available at: http://ic.shu.edu.tw/present/85/LibAuto/D1F.htm. Accessed Apr.30, 2003.
62.羅友謙，XML-Based電子型錄互通性先導計畫---XML之特色和現況概述；Available at: http://www.nii.org.tw/cnt/Project/xml_based/cat_project2.htm. Accessed Jul. 23 2005.
62.MEDLINE醫學文獻資料庫.Available at: http://www.hint.org.tw/dbguide/medline.htm2030216. Accessed Apr.30,2003.
67.張慧銖(2000)，統一醫學語言系統簡介；台大醫學院圖書分館館訊第廿七期第1-4頁Available at: http://www.lib.ntu.edu.tw/pub/mk/mk27/mk27_1.html。Accessed Jun. 25, 2005.
69.Sun Microsystems(2002). "The Java Language: A White Paper", Available at: http://java.sun.com/doc/overview/index.htm. Accessed Feb.16, 2003.
84.MEDLINE全科性醫藥學文獻檢索資料庫；Available at: med.tajen.edu.tw/DB/Medline. Accessed Feb.16, 2003.
85.全科性醫藥學文獻檢索資料庫；Available at: http://med.tajen.edu.tw/dbmenu/dbintro.htm Accessed Jul. 23, 2005.
91.Louis YK, Bidgood, Jr. (1997). Representation of the Gastrointestinal Endoscopy Minimal Standard Terminology in the SNOMED DICOM Microglossary。Available at: www.amia.org/pubs/symposia/D004219.PDF. Accessed Feb.16, 2003.
97.阮相宇，翁昭旼，自然語言處理在醫學資料庫系統上的應用.台灣大學資訊工程系所「自然語言處理實驗室」。Available at: http://www.datamining.org.tw/doc/paper_data_detail.asp?no=89 Accessed Jun.25, 2003.
113.Crangle C, Zbyslaw A, Conver Speech LLC M. Cherry, E. Hong, Stanford University: Concept Extraction and Synonymy Management for Biomedical Information Retrieval. Available at: http://trec.nist.gov/pubs/trec13/papers/converspeech.geo.pdf Accessed Jul. 23, 2005.