ASTRACT: The evolution of the Semantic Web has accelerated the need for ontologies. There is a need to build new ontologies and to extend and merge existing ontologies. To achieve this we need software tools to edit and build ontologies. Our paper describes software to develop a protocol for collaborative ontology editing based on RDF and using a Peer-to-Peer (P2P) networking architecture. The protocol allows for the implementation a voting mechanism embedded into the RDF data itself, using a mixed initiative design for notification. This is implemented as extensions to an ontology browser called ONTORAMA (1). The P2P approach is compared to the classic ontology editing approaches and the special requirements of the ontology editing environment are discussed. The protocol, design, implementation and architecture for ontology update are also elaborated.
Categories and Subject Descriptors
C.2.1 [Network Architecture and Design]; D.2.12 [Interoperability Web-based Services]
Web services, Peer to peer network, Ontology
Keywords: Semantic web, Ontology, P2P Network architecture, RDF data, Network protocol
The Semantic Web  is a vision of the Web where available data can be processed automatically by machines. In this vision, machines are able to retrieve and process data to make meaningful decisions and agents are able to communicate and negotiate to achieve sophisticated tasks set by users. To achieve this RDF was developed as a language that records sets of triples to describe any resource accessible on the Web. The power of RDF is that it maps the representation directly into a data model. This data model is simple but with a number of syntactic variations, most noticeably designed around an accommodation of XML syntax.
One of the main reasons for the appearance of RDF within the context of the Semantic Web is for the creation of ontologies--explicit specifications of conceptualizations specific to an application domain.
RDF gives an infrastructure to model such ontologies, which can be used to communicate available terminology to a user of a website or to enable semantic inter-operation between autonomous agents.
The paper is structured as follows. First, we present a literature survey that highlights ontology server tools and the difficulties associated with a client-server architecture for ontology management. Second, we elaborate on the task of ontology editing, what research in this area has achieved and how our design is adapted from previous successes in distributed collaboration. Second, we present an overview of P2P systems. Many of the features for sharing ontological content derive from other successful P2P file sharing systems. We characterize our work in the context of the feature space of well-known P2P systems with the additional (and special) requirement of ontology editing. Third, the high-level mechanisms for collaborative P2P ontology editing are developed. The idea of voting, via assertions and rejections on content, gives a mechanism to control ontology update and a principled approach to knowledge-based version control management. Since our solution is embedded within the ONTORAMA ontology browser, the fourth section of this paper describes ONTORAMA, its capabilities and limitations. We spent some time in the fifth section describing the implementation details: demonstrating the operation of the P2P editor in use cases. In future work, we will stress test the application to explore the bounds of its practical use as a distributed P2P system.
1. Background and Literature Survey
Our paper presents a P2P ontology editing environment based on the ontology browsing tool ONTORAMA. The prototype, derived from ONTORAMA, provides a basic editing capability and ontology visualization framework. Further, a P2P network protocol facilitates collaborative ontology editing. The environment is decentralized and distributed. We used DAML+OIL compliant RDF to express our ontologies which enables interoperability with other ontology tools using the same standard.
Ontologies are normally maintained as static text files or by using a centralized server/client approach. The former is suitable for small ontologies that do not require frequent update but not particularly suitable for collaborative work. Text-based version control systems can be used to manage updates but their use for highly structured data like ontologies is limited. On the other hand, a client/server approach to managing ontological content enables collaboration but comes with technical and administrative overhead.
The aim of having a single large, on-line ontology of semantic data is realistically unachievable, so smaller (often domain specific) Web-based knowledge bases have emerged (see Ding ). As a result of distributing metadata descriptions identifying resources and their properties can be problematic. Different systems or databases often use different terminologies to describe the same object or idea. Alternatively, a single descriptor can be associated with a number of different objects or ideas. Many ontologies also allow the definition of inference rules that allow conclusions be drawn, conceptualizing the data encountered into a domain context and in so doing relating it to the task at hand. Ontologies have other uses: (i) as a common vocabulary for people working in the same …