Article Preview
TopRecent development of semantic Web has fuelled researches on developing semantic search engine and methods to exploit the potential of semantic Web, particularly the ontology, in information retrieval. Vallet et al. (2005) combine semantic search with keyword-based search. Ontology-based scheme is used to annotate documents. Weighted average is used to combine the semantic similarity measure with the similarity measure derived from the classic vector-space model. Dong et al. (2008) presents a preliminary survey on semantic search technologies and discusses common issues in the current semantic search engines and methods. Semantic search approaches are classified into various categories including semantic search engines and semantic search methods that completely adopt the semantic Web technology, hybrid semantic search engines that integrate semantic web technology into key-word-based search engines to improve the precision of traditional text search, search engines that can query objects in XML documents, and search engines that are designed for querying ontological files. Major common issues of semantic search methodologies include differentiation between designers and users’ perceptions of relevancy of context, lack of adaptability in the knowledge structure, and lack of experimental tests to verify the model.
Mangold (2007) also studied 22 semantic document retrieval systems and proposed a categorization scheme based on seven features: architecture, coupling, transparency, user context, query modification, ontology structure and ontology technology.
Architecture concerns whether the underlying design of system stores an index of documents. A system is considered as a stand-alone search engine if it maintains indexes of the document data. Examples include SHOE (Heflin & Hendler, 2000) and SCORE (Sheth et al. 2002). A system is called a Meta search engine if it sends queries to subordinate search-engines. For instance, Inquirus2 (Glover et al., 2001) and TAP (Guha et al., 2003) employs the idea of Meta search engine.
Coupling is about how close the relationship between the documents and ontologies. In tight coupling, the ontology concepts are closely related to the documents. For instance, both hybrid spreading activation (Rocha et al., 2004) and librarian agent (Stojanovic, 2003) adopts tight coupling approach. On the contrary, the loose coupling refers to the case that documents are not committed to any ontology. Both ISRA (Burton-Jones et al., 2003) and TAP (Guha et al., 2003) make use of loose coupling.