The usability of semantic search tools: a review

VICTORIA UREN, YUANGUI LEI, VANESSA LOPEZ, HAIMING LIU, ENRICO MOTTA, MARINA GIORDANINO
2007 Knowledge engineering review (Print)  
The goal of semantic search is to improve on traditional search methods by exploiting the semantic metadata. In this paper, we argue that supporting iterative and exploratory search modes is important to the usability of all search systems. We also identify the types of semantic queries the users need to make, the issues concerning the search environment and the problems that are intrinsic to semantic search in particular. We then review the four modes of user interaction in existing semantic
more » ... arch systems, namely keyword-based, form-based, view-based and natural language-based systems. Future development should focus on multimodal search systems, which exploit the advantages of more than one mode of interaction, and on developing the search systems that can search heterogeneous semantic metadata on the open semantic Web. Introduction When cannot a seat be sat upon? One answer to this riddle might be: when you search for it on the Web. In most of the popular Web search engines, such as Google, IceRocket, ASK and AltaVista, the top hits on a search for 'seat' are about SEAT, the Spanish car maker. It is problems like this with keyword-based search that are driving the quest for semantic search engines. The idea is that a semantic search engine would support the user in formulating a search to get, not the results about SEAT cars or seats of government but the results about furniture you can sit on, like chairs, sofas and benches. In this review, we explore the types of features semantic search systems require, and take a look at some of the semantic search systems that have been developed so far. The phrase 'semantic search' is used to describe several different kinds of search. To clarify the kind of semantic search discussed here, we are not interested in searching for ontologies, as is done by systems like Swoogle (Ding et al., 2004) . Nor are we interested in semantic document search, which employs a combination of semantic metadata and other document features to enhance the performance of document retrieval, for example, Vallet et al., (2005) . Our interest lies solely in searching the semantic metadata. We do not constrain ourselves to any particular ontology language, such as OWL Full, Resource Description Framework Schema (RDFS), rather the scope of our review is the search systems that operate on RDF triples and which may contain uniform resource identifiers (URIs) that link to Web-accessible resources (e.g. 'Enrico Motta' has homepage http://kmi. open.ac.uk/people/motta/). We refer to such RDF triples as 'semantic metadata'. We define an annotation as an RDF triple, which is linked to a Web resource and represents the fact that a particular part of a resource denotes a particular concept. Annotations can be understood to include the relevant fragment of the original resource, but we concern ourselves only with the basic RDF triple. Therefore, we define metadata-oriented semantic search as the search which: * enables the querying of RDF triples, both semantic metadata and annotations; * enables access to Web resources referenced as URIs in RDF triples; * works on semantic metadata repositories, often called as knowledge bases or KBs, specifically RDF KBs, as used by the semantic Web; * and makes use of semantics defined by domain ontologies to support query formulation and to improve the precision of search results.
doi:10.1017/s0269888907001233 fatcat:seelzyyscngcpbymqaunefzly4