專家診斷系統應用於文書處理術科評量機制之研究

Abstract

本研究之主要目的為降低文書處理術科檢定試題中，人工閱卷之繁 瑣性、困難度及可能產生之誤差，同時提高診斷結果的正確性及效率，從研究中創造知識經濟的實際效益及專家系統之價值；而為達到此一目的，本研究於過程中需尋找適當之搜尋法及演算規則，並透過實際制定專家知識庫及規則庫、分析推理引擎與各單元之間的運作情形來撰寫此一系統。 研究方法以專家系統為理論基礎，分析市面上現有之檢定機構所發 行的術科(performance-based)試卷，發展出一套以專家經驗為評量標準的電腦術科測驗系統，其過程透過了分解(Resolution)、資料挖掘(Data mining)、實體關係資料模型(E-R Model)及派翠西網路(Perti-Nets)等基本理論的運用與分析，以及提出CAI 檢定及測驗軟體專屬之S-O-A-V(Score-Object-Attribute-Value)圖作為理論的依據及系統發展的基礎。 依據S-O-A-V 圖形及演算規則，本研究分析設計出專家規則知識庫 及程式演算規則，最後設計出ActiveX 元件及測驗網頁，並委請多位教師於教學上使用，給予學生實際測驗；測驗結果顯示人工閱卷及系統閱卷兩者間僅有極小之差異性存在，且系統於再次修正後，確實能提高準確性、正確率，及克服人工閱卷可能產生之缺點，對於教師教學、學生練習及檢定工作有著極大的助益，證實此一模式值得推廣採用。

The purpose of this research is to avoid human’s complicated, difficult and possible error in checking performance-based exam, and to increase the correct rate and efficiency of the diagnosing result, it can create the real benefit of knowledge economic and the value of Expert System; For reaching that, researcher needs to look for suitable searching methods and rules to establish then expert knowledge based, the rules based, to analyze the inference engine’s operation between each unit, then write this system.

The method of this research is based on the theory of Expert System. It analyzes the performance-based testing papers existing and had issued by the examining organizations on the business situations. Then developed a performance-based testing system based on the experts’ experiences. This system is by way of using and analyzing some basic theories, such as resolution, data mining, entity-relation model and perti-nets, and bring out CAI and the S-O-A-V (Score-Object-Attribute- Value) figure, which is using only in examining and testing softwares, to be the base of the theory and the system development.

According to the figure of S-O-A-V, this research brings out the warehouse of specialist rules and knowledge and the calculation rules of composing formulas. Finally, it designed ActiveX components and the testing web pages. These web pages had been used by many teachers in instruction, and had been tested by many students practically. The testing result had displayed a few differences between human and system, and it had raised accuracy and correct rate after re-correcting, it can overcome human’s fault, and has a lot of benefits on teacher’s instruction, student’s practicing, and examines, it proves it is worthy of popularizing and using this mode.