隨著資訊科技的日新月異，軟體市場逐漸擴增，而為了在軟體市場中競爭，除了開發時程要符合客戶需求外，軟體品質也是一項至關重要的要素。倘若客戶要求的時程過短，開發人員會選擇較彈性或敏捷的開發方式，進而省略了事前的系統設計文件，直接進行軟體實作。然而，系統的設計文件最主要的兩個用途是在軟體實作階段確保人員撰寫出品質好的系統，以及在維護階段幫助人員理解系統內容。而目前最常見的做法是利用UML反向工程工具於系統完成後產生與系統相對應的UML圖形，但是這樣的做法卻帶來了潛在的品質問題。若能有一套機制可於UML圖形產出後，進一步探討該圖形之品質問題，則能有效協助人員修正系統品質。 因此，本研究針對UML圖形中的結構模型圖之品質進行探討，並以耦合度的靜態分析與動態分析作為品質評核之標準。此外本研究結合了本體論方法來建立品質評核之知識架構，使領域知識能隨著時間與地點的不同進行修正。結合上述兩點，本研究實作了一套Web-based的UML圖形的品質評核系統－Ontology-based Reverse Engineering Diagram Quality Inference System，並透過一個Java專案來展示該系統之功能與效益，最後以專家訪談的方式驗證本系統。;Not only satisfying clients’ need in time but also delivering quality software is the major requirement in software market along with the development of IT industry and the boom in software developing environment. Some projects might choose a more agile or flexible approach of development in order to fulfill the requirement with less time by skipping the process of creating documents for the project beforehand. However, the two main purposes of foregoing documents are assuring the quality of software and helping personnel to understand how system works. One of the most common ways to create UML diagram for agile projects after software development is to use UML reverse engineering tools. But UML diagrams generated by UML reverse engineering tools have potential quality problems. This study tried to establish a model containing multiple measures for evaluating the quality of structural UML diagram created by UML reverse engineering tools. These measures are coupling metrics measure by static analysis and dynamic analysis and using ontology to help for inferring quality rule. In addition, this study propose an evaluation system－Ontology-based Reverse Engineering Diagram Quality Inference System(OntREDQ), which aims at evaluating quality of UML diagram using foregoing model. The result of evaluation would be demonstrated by a Java project and the effectiveness of the model would be verified and discussed by professional experts.
