以本體論為基礎具半結構化及演化特性之資訊服務點對點網路

Abstract

本體論已成為資訊管理的新典範，而點對點系統則為分散式資訊共享提供了一個彈性的網路環境。由於結合本體論與點對點系統可以達到綜效，許多學者都開始研究如何將它們適當地結合。其中，有些研究是奠基在 DHT 網路之上。然而，DHT 的本質可能招致一些無法避免的缺點，如難以支援豐富語意的查詢、在分散的節點上維護 DHT 而導致負荷、以及對節點自主性產生限制等等。相對的，有些研究則運用中央化機制，以建構一個支援彈性且豐富查詢的網路。但是，使用中央化機制將會使系統遭受一些先天上的問題，如易受攻擊和當機之患。以上所提到這幾點，在在都突顯出我們需要一個同時兼顧彈性與強固性的設計。 我們運用嶄新的方法，構思出一個發揮本體論與點對點系統之綜效的網路，稱作『OntoZilla』。在這個系統中，每個節點上的資源，都是利用本體論技術來作管理，因此這個網路中的各種資訊都能被自動化處理。另一方面，我們也運用本體論來組織節點。藉由使用“語意連結”來反應節點之間的關係，構成了一個以本體論為基礎且具半結構化特性的網路拓撲。為了使節點之間的關係能保持最新，節點們彼此會不斷地“交換訊息”，使他們的關係得以持續演化。再者，我們也針對容錯、負載平衡、穩定化等方面提出對策，使我們的系統更能自我管理並適應各種情況。這個智慧的系統可以被運用到許多領域，如可以當做概念檢索的平台，或是動態服務探索與整合的平台。我們相信這個系統將可以改善資訊檢索的流程，並且促進更佳的系統相互整合與溝通。

Ontologies are a new paradigm for information management, while peer-to-peer (P2P) systems provide a flexible network infrastructure for distributed information sharing. As combining ontologies and P2P systems would result in synergy, many researchers have focused on a combination of them. Some research is based on Distributed Hash Table (DHT) networks. However, the nature of DHTs may necessitate some compromises, because of the difficulty in supporting expressive queries, the overhead of maintaining a hash table at distributed nodes, and the restrictions on peer autonomy. In contrast, other research leverages centralized mechanisms to build a network that supports flexible and expressive queries. However, the use of centralized mechanisms makes systems vulnerable to some inherent problems, such as attacks and failures. All these drawbacks highlight the need for a robust design that also meets the requirement of flexibility. Taking a different approach to the above, we propose a system called “OntoZilla”, which utilizes the synergy between ontologies and P2P systems. In our system, resources at peers are managed using ontologies, thereby enabling automatic processing of information in the network. On the other hand, we exploit ontologies to organize peers. Using “semantic links”, the relationships among peers are reflected, resulting in an ontology-based and semi-structured network topology. To keep peer relationships up-to-date, peers constantly “gossip” with each other so that their relationships evolve. Furthermore, we tackle the challenges of fault tolerance, load balance, and stabilization so that the system is self-supervising and adaptive. This intelligent system can be applied to many areas. For example, it can be used as a platform for concept searches, or for dynamic service discovery and integration. We believe our system would improve the process of information searching and facilitate greater integration as well as interoperability.