MOVE: A distributed framework for materialized ontology view extraction

The use of ontologies lies at the very heart of the newly emerging era of semantic web. Ontologies provide a shared conceptualization of some domain that may be communicated between people and application systems. As information on the web increases significantly in size, web ontologies also tend to grow bigger, to such an extent that they become too large to be used in their entirety by any single application. Moreover, because of the size of the original ontology, the process of repeatedly iterating the millions of nodes and relationships to form an optimized sub-ontology becomes very computationally extensive. Therefore, it is imperative that parallel and distributed computing techniques be utilized to implement the extraction process. These problems have stimulated our work in the area of sub-ontology extraction where each user may extract optimized sub-ontologies from an existing base ontology. The extraction process consists of a number of independent optimization schemes that cover various aspects of the optimization process, such as ensuring consistency of the user-specified requirements for the sub-ontology, ensuring semantic completeness of the sub-ontology, etc. Sub-ontologies are valid independent ontologies, known as materialized ontologies, that are specifically extracted to meet certain needs. Our proposed and implemented framework for the extraction process, referred to as Materialized Ontology View Extractor (MOVE), has addressed this problem by proposing a distributed architecture for the extraction/optimization of a sub-ontology from a large-scale base ontology. We utilize coarse-grained data-level parallelism inherent in the problem domain. Such an architecture serves two purposes: (a) facilitates the utilization of a cluster environment typical in business organizations, which is in line with our envisaged application of the proposed system, and (b) enhances the performance of the computationally extensive extraction process when dealing with massively sized realistic ontologies. As ontologies are currently widely used, our proposed approach for distributed ontology extraction will play an important role in improving the efficiency of ontology-based information retrieval.