Exploiting metadata links to support information retrieval in document management systems

Abstract

A hypothesis of this dissertation is that utilising metadata linked into the ontological structure for the tasks of supporting information retrieval, offers an advantage over the traditional full text searching techniques. The "à la" (Associative Linking of Attributes) concept presented in this thesis demonstrates how a system for managing document collections in an organisational setting, could be enhanced to support the information delivery process. It can be considered as a pre-Semantic Web application, a recommender facility that provides assistance to locating items by utilising hypertextually linked metadata. The contribution of this thesis lies in three areas:

•     Novel approach to metadata linking: using ontological zzstructures and ZigZag linking;

•     Novel approach to using the metadata network for document retrieval: "Query by association";

•     Evaluation that compares the usage of the " à la" prototype system with the traditional approach.

The "a la" prototype evaluation study compares free text querying using a full text search approach with the "a la" method for finding relevant documents. Precision (fraction of relevant search results) and serendipity (fraction of novelty or positively surprising items in a search result) were the chosen metrics. The study findings indicate that " à la" performs better in carrying out general-concept queries that fall within a project's knowledge domain. The evaluation concludes that " à la" can successfully enrich a typical document management system in order to improve the user's searching experience, by bringing not only the expected relevant search results, but also the items which are serendipitous. Moreover, browsing metadata connected in a Zigzag fashion can better aid understanding of the existing domain. This can in turn benefit knowledge engineers in their first step towards building a more formal organisational ontology. It is concluded that the " à la" method has the potential to be applicable towards establishing principles of the Semantic Web within a corporate environment.

More information

Identifiers

Catalogue record

Export record

ASCII CitationAtomBibTeXData Cite XMLDublin CoreDublin CoreEP3 XMLEndNoteHTML CitationHTML CitationHTML ListJSONMETSMODSMPEG-21 DIDLOpenURL ContextObjectOpenURL ContextObject in SpanRDF+N-TriplesRDF+N3RDF+XMLRIOXX2 XMLReferReference ManagerSimple Metadata