UMBC ebiquity

Finding and Ranking Knowledge on the Semantic Web

Authors: Li Ding, Rong Pan, Tim Finin, Anupam Joshi, Yun Peng, and Pranam Kolari

Book Title: Proceedings of the 4th International Semantic Web Conference

Date: November 07, 2005

Abstract: Swoogle is a system that helps knowledge engineers and software agents find knowledge on the web encoded in the semantic web languages RDF and OWL. Based on the search mechanisms provided in the previous version, we propose a novel semantic web navigation model and refine mechanisms for ranking the semantic web at various granularities. Although the semantic web is materialized on the Web, it is hard to navigate within the semantic web since few explicit ``hyperlinks'' are available besides a URIref's namespace or owl:import semantic. Hence we propose a navigation model that characterizes users' navigational behavior (e.g. surfing from an ontology to one class C defined in it, and then to the RDF documents that populate C or the other resources that help defining this class) within the semantic web and implement it in Swoogle's ``Ontology Dictionary''. Based on this navigation model and the metadata collected in Swoogle, we have developed algorithms for ranking objects in the semantic web at various levels of granularity including semantic web document (SWD) level, resource level (e.g., RDF class or property) and triple level (e.g. interesting RDF graph pattern). Ranking SWDs, inspired by the Google's PageRank, emulates an ``rational'' agent acquiring knowledge on the semantic web using the hyperlinks provided by our ``semantic web navigation model'' at document level. Ranking individual terms extends ranking to a finer granularity. For example, from the hundreds of RDF terms denoting the concept of a person, the question of ``which are most widely used?'' is answered by term ranking. Finally, we introduce the notion of ranking facts (e.g., RDF triples) such as the rdfs:domain relation between a class and a property using provenance based heuristics. These ranking mechanisms, if being used in filtering ontologies, could help the emergence of consensus ontologies. Experiments show that the Swoogle search engine using ``semantic ranking'' outperforms Google in evaluating the importance of ontologies.

Type: InProceedings

Publisher: Springer

Series: LNCS 3729

Pages: 156--170

Tags: semantic web, swoogle, information retrieval, rdf, owl

Google Scholar: MN7NZvxbsM8J

Number of Google Scholar citations: 130 [show citations]

Number of downloads: 11610


Available for download as

size: 363476 bytes

Related Projects:

Past Projects

 Semantic Discovery: Discovering Complex Relationships in Semantic Web.