| Previous | [ 1] | [ 2] | [ 3] | [ 4] | [ 5] | [ 6] | [ 7] | [ 8] | [ 9] | [ 10] | [ 11] | [ 12] | [ 13] | [ 14] | [ 15] | [ 16] |
¡@
Fidel Cacheda, Victor Carneiro, Carmen Guerrero and Angel Vina
Department of Information and Communications Technologies
University of A Coruna
A Coruna, 15071 Spain
Search systems based on hierarchical taxonomies provide a specific type of search
functionality that is not provided by conventional search engines. For instance, using a
taxonomy, the user can look for documents related to just one of the categories of the
taxonomy. This paper describes a hybrid data architecture that improves the performance
of restricted searches for a few categories of a taxonomy. The proposed architecture is
based on a hybrid data structure composed of an inverted file with multiple integrated
signature files. A detailed analysis of superimposing codes on directed acyclic graphs
proves that they adapt perfectly well to a search system based on a hierarchical ontology.
Two variants are presented: the hybrid architecture with complete information and the
hybrid architecture with partial information. The validity of this hybrid architecture was
analyzed by developing and comparing it with a basic architecture. The performance of
restricted queries is clearly improved, especially with the hybrid architecture with partial
information. This variant outperformed by 50 % the basic architecture for all workload
environments, with a slight reduction in performance for the lower levels of the graph.
Received April 13, 2004; revised August 11 & December 27, 2004; accepted January 20, 2005.
Communicated by Ming-Syan Chen.
* This paper was partially supported by the CICYT of the Spanish government, under project TIC2001-0547.