?服務搜尋為服務導向領域相當重要的核心技術，過去以關鍵字比對和服務分類為主要之搜尋方法並無法有效地幫助使用者找到他們所需要的服務，而使用者標註(User Annotation)或知識本體(Ontology)等方法需花費大量的額外人力方可增添詮釋資料(Meta Data)。? 本論文之目標在於建構一個服務搜尋引擎，透過剖析標準之網際服務描述(Web Services Description Language，WSDL)以建構本論文所定義的服務摘要(service signature)圖形，接著搭配語彙分析、資料型態比對等方法計算出使用者需求與各網際服務的相似度，以擷取出最接近使用者想要的網際服務。本論文並提出多種提升比對運算效率與提升配對準確性之方法，實驗證實本論文提出之方法確實具有顯著之成效。? Web service discovery is an important technology in field of service- orientation. Nowadays, keyword-based and category-based service discovery methods are not effectively finding appropriate services for users. Meanwhile, service discovery approaches using techniques of user annotation or “ontology are usually costly since service providers and service requesters need to spend considerable effort to add semantic annotation on service descriptions and service requests.? In this study, we proposed a graph-based web service matchmaking approach based on the open specification WSDL (Web Services Description Language) and accordingly developed a web service search engine. The proposed approach extracts significant elements from a WSDL document and constructs Service Signature Graph to represent the web service, and performs lexical analysis and data type comparison to calculate the similarity degree between the user’s request and all candidate services to help users to precisely identify appropriate web services. In addition, we proposed several methods to improve the precision of search results and the performance of the discovery process. The experimental result shows that the proposed methods are effective to improve the search precision and the execution performance.
