以階層式案例推論方法輔助建築平面考試試題設計

Abstract

隨著電腦及網路科技的進步，藉由電腦及網路的學習評量稱為CBT (Computer-Based Testing)已成為一個趨勢，對美國建築師考試(ARE， Architect Registration Examination)而言，電腦化測驗已普遍應用在考試的各個項目，其中以圖形化(Vignette)測驗佔大多數，目的在藉由圖面的繪製來測驗出知識整合的能力，而在本研究中，我們將專注於建築平面(Schematic Design)設計的項目。在試題的出題設計上因為是測驗高階的設計知識概念，所以在出題上對老師是很困難的。他們通常需要去做複雜且多層次的考量，定義且設計試題，以確保能有效的測驗出學生的能力，所以常常會花很多的時間及成本。我們的目標是建立一個輔助出題系統去協助老師出題且能對之前的考題再利用，讓出題不再是從無到有那麼的困難。目前考題的再利用方面，往往是圖面的蒐集整理，沒有結構化的去對圖面內容做定義，以致於再利用的程度有限。所以如何結構化的描述建築平面設計知識模型是一個很重要的課題，而模型的描述能力、延伸性及操作能力是一個很重要的評估指標。如何清楚的描述知識的特性，我們的概念是藉由專家的角度及電腦科學的技術來設計我們的模型，HCBR(Hierarchical Case-Based Reasoning)的方法是提出一個階層性的概念，來對知識做有層次的分類，以案例為基本單位加以推論及再利用。除此之外，我們提出一個智慧型的搜尋條件產生器(Intelligent Query Generator ，IQG)，IQG 能有效協助老師對系統下搜尋條件，藉由搜尋條件的增減的建議，能找到適合的案例，搜尋條件可當成考題，而搜尋出來的案例可以當成改考卷的參考答案；其中我們所定義的搜尋語言(Query language， QL)是協助IQG運作的重要角色，目的在有效描述輸入的條件。最後，我們會提出系統設計上的建議，在實作方面，我們會實作出建築平面比較器，並以問卷的方式，來驗證系統的成效。

With the growth of the computer and the Internet technology, the learning assessment via the computer and the Internet, called Computer-Based Testing (CBT) has become a trend. In the Architect Registration Examination (ARE) the CBT is applied in schematic design test item to assess students planning capability. In this paper we focus on schematic design test item of subject building planning in ARE. However, the design of test item for assessing high level knowledge is difficult for teachers. Since they usually need to concern complex situations to make sure the effectiveness of test, the creation of test item is usually time consuming and costly. Our goal is to construct a system to assist teachers in item generation by reusing desirable cases. However, the database of current design cases usually focuses on documentary collection for layout information and is ill-structure for design knowledge representation. Therefore, how to organize the design knowledge to represent multidiscipline and granularity property of building for reusing is an interesting and important issue. The expressive power, extensibility and manipulation of model are also our issue. To clearly represent the building properties for above issue, our ideal is to organize design attributes form experts' perspective by extending the layout with environmental context information. Accordingly, the Hierarchical Case-Based Reasoning (HCBR) approach is proposed to organize the knowledge granularities by hierarchical relations. Therefore, the Intelligent Query Generator (IQG) is proposed to support the verification of constraint rule during item design by iterative case retrieval process. If the designed constraint rules are too specific which means no case is retrieved, then the IQG will recommend the user release the constraint rules. With the IQG the teacher can easily generate the desirable test item. To evaluate the effectiveness and usability of proposed approach, the prototype system and ten test cases were implemented. The corresponding questionnaire analysis was applied and the result shows that the proposed approach has high satisfaction degree.