The Semantic Web is characterised by thousands of ontologies and millions of Semantic Web documents. Ontology Search Engines, such as Swoogle and Watson, crawl and index this information, to provide human users and applications with information about what is available on the Semantic Web. However both Watson and Swoogle only deal with content which is publicly available and made available through a centralised gateway. In the OpenKnowledge project we have also looked at modelling “Cautious Knowledge Sharing” (CKS) scenarios, where knowledge providers retain control about which knowledge is exposed and provided to consumers. In particular we have developed an infrastructure for exploring CKS scenarios by distributing the Watson Semantic Web Gateway over a network of OpenKnowledge peers. In order to support ‘cautious knowledge adverts’ we have also developed a groundbreaking algorithm, which summarises an ontology in terms of a small number of key concepts. The results generated by this algorithm have been shown to correlate with those produced by human experts. We have also carried out a comparative evaluation of our infrastructure for CKS with the centralized version of Watson, in the context of the PowerAqua query answering system. The evaluation has shown that the use of automatic key concept extraction has negligible impact on recall, while spreading ontologies over a network of Watson peers improves the overall efficiency of the system.
More information is provided in Deliverables 8.4 and 8.5.