支援決策知識庫之文件資料自動擷取與分析之研究

Abstract

在發展專家系統的過程中，如何建構一個完善的知識庫，是一項值得研究 的課題。本論文以發展豬隻疾病診斷之專家系統為例，在台灣養豬科學研 究所內，設有一研究用途的豬場，自從1988年後，對豬場內豬隻疾病的診 斷，已留下豐富的病歷資料，這些病歷資料對我們所要發展的專家系統的 知識庫部份而言，將是一些寶貴的原始資料來源。這些病歷是以文件的格 式儲存，傳統的方式，是透過專家的協助，以人工的方式，來將這些原始 資料整理成專家系統可使用的知識庫，如此的方式將耗費相當多的人力資 源。如果能將這部份的工作以電腦來取代，不只可減少所需的人力，對整 個知識庫建構成本的降低，亦有相當大的幫助。本研究建立了一套原始文 件資料的自動擷取系統，透過這個系統能將這些原始的文件資料，建立成 一個病歷的資料庫。此資料庫將能透過專家系統的學習機構，轉換成專家 系統所需之知識庫，如此便能有效利用這些豐富的病歷資源，供相關領域 的研究者或業者使用。 In the process of developing the expert system, the more valuable part is to construct an ideal knowledge base. In the thesis, we take the diagnostic system of swine disease as an example of an expert system. After 1988, there are plentiful diagnosis's cases of swine disease from an experimental pigpen in the Institute of Swine Breeding in Taiwan. These cases are important resources of the database by which we will build an expert system. In a traditional way, we have to construct a knowledge in the directions of the experts by inputting the cases artificially, and then we build an expert system by inputting those original data. This way will take a lot waste of work force. If we can replace the part of work with computer, we will not only decrease the need of personnel, but also reduce the total cost of building the knowledge base. In my research, we have built a system to generate knowledge base from documents automatically. We can make those original source files become a database of case history by using this system. Then by the learning department of the expert system, we can transform the database into a necessary knowledge base for the expert system. Therefore, we can use all of the abundant cases' resources effectively, providing for the researchers and business persons.