運用知識本體與法則之網路服務近似媒合與評等

Abstract

網路服務，一個支援在網際網路上電腦跟電腦間互動的軟體架構，已經將目前的網路從一個靜態的資訊聚集地轉型成為一個動態的服務互動媒介。隨著網路服務的數量日益成長，服務媒合遂成為了一個越來越重要的課題。服務的媒合主要是指針對使用者所提出來的需求，透過搜尋線上的服務目錄，而找到所有滿足需求的服務廣告並將之回傳給使用者的一個過程。然而，由於每個人的對於其所期望的服務需求是因人而異的，我們很難去預期每次使用者透過服務媒合機制所找到的服務都會完美地符合他們的需求。因此，當無法搜尋到完美符合的結果時，我們就需要有一個近似媒合的機制，它能夠回傳能近似滿足使用者需求的替代服務並提供評等的結果給使用者參考。 在本論文中，我們整合之前所提出來能處理「量化數值關係」以及支援「媒合精確度設定」的媒合方法，而延伸前述的方法而加入了一個在描述邏輯裡表達「概念組合」的方式，概念組合主要是用於當我們要使用一個概念表達透過某個屬性來連結的兩個概念，由於在描述邏輯裡並不支援這種建構子，因此我們將每一個位於子節點的概念用一個個體來表達，我們利用一個特別建立的屬性來表達概念間的組合關係。在我們的方法裡，領域的知識是透過知識本體跟法則的形式存放在知識庫中， 我們利用描述邏輯裡的概念來代表服務的描述，透過這種方式，我們可以將服務媒合的問題轉化成為驗證概念跟概念間的包含關係。為了能夠近似服務的描述，我們將需求描述裡面的限制逐步放寬，放寬的方法主要是將每個條件用更廣義的概念來取代。我們將搜尋結果透過評等方法評定相對優劣等級提供給使用者使他們能透過評等結果來決定最適合的服務。為了驗證我們的方法，我們開發了一個基於旅遊領域的雛型系統－媒合者，這個媒合系統能夠被完全地整合到「旅行者」這個服務組裝的架構，這樣子的一個整合架構可以為我們展示從服務描述取得、服務媒合到服務執行的一個完整的過程。透過我們的系統，使用者能夠描述他們針對特定領域限制條件的關切程度並獲得近似符合他們需求的替代性服務。

Web Service technology, which is a software framework supporting interoperable machine-to-machine interaction over the Internet, has transformed the Web from a static collection of information into a dynamic service repository. As the number of available Web services continues to grow, service matching becomes an increasingly important issue. Matching is the process that requires a service repository to take a requirement as input, and return all the advertisements that satisfy the requirements specified in the input query. However, since users' requirements may vary from person to person, it is unrealistic to expect that a matching scheme will always find services that match requirements perfectly. When exact matches cannot be achieved, an approximate matching scheme, which gives satisfying and ranked results, will be needed. In this thesis, we integrate previous approaches that can deal with 'quantitative relations' and allow users to specify 'matching accuracy settings'. We then extend them with an approach to model 'concept composition'. Concept composition could be useful when we want to compose two concepts linked by a certain property. Since there is no such constructor in Description Logic, we use individuals to represent concepts which are leaf nodes of the ontology. Then, we use a role to represent the composition relationship. In our scheme, domain knowledge is stored in the knowledge base in the form of ontologies and rules. We model service descriptions as concept expressions. Therefore, service matching problems can be transformed into concept subsumption checking, which is a basic reasoning service supported by reasoners. To approximate users' requirements, we try to relax the constraints of requirements by substituting these constraints with more general ones. Furthermore, we rank these results according to our ranking scheme, which can provide a relative priority for users to select services that best fit their needs. To validate our approach, we implement a prototype system - the Matchmaker, which focuses on the tourism domain. This system is fully integrated with the Traveller, which is a prototype system of the service composition framework that demonstrates the process from service description acquisition, service matching to service execution. With our matching system, users are able to specify their concern toward certain domain specific constraints by matching accuracy settings and acquire substitute services that approximately fit their needs.